From bc35fd859e89a1e2ff009fedeb6f59b4cfc26656 Mon Sep 17 00:00:00 2001 From: sglang-bot Date: Mon, 16 Nov 2026 00:00:30 +0000 Subject: [PATCH] chore: update CI test est_time from recent run data --- test/registered/attention/test_aiter_gluon_h12_fp8.py | 2 +- test/registered/attention/test_create_kvindices.py | 2 +- test/registered/attention/test_hybrid_attn_backend.py | 2 +- test/registered/attention/test_qwen35_deterministic.py | 2 +- test/registered/attention/test_triton_attention_kernels.py | 2 +- test/registered/attention/test_trtllm_mha_encoder_only.py | 2 +- test/registered/attention/test_trtllm_mha_page_table.py | 2 +- test/registered/attention/test_verify_splitkv.py | 2 +- .../attention/unittests/dense/test_extend_init_contract.py | 2 +- test/registered/attention/unittests/dense/test_fa3.py | 2 +- test/registered/attention/unittests/dense/test_fa4.py | 4 ++-- .../registered/attention/unittests/dense/test_flashinfer.py | 4 ++-- .../attention/unittests/dense/test_flex_attention.py | 2 +- .../attention/unittests/dense/test_hybrid_attn.py | 4 ++-- test/registered/attention/unittests/dense/test_tbo.py | 4 ++-- .../attention/unittests/dense/test_torch_native.py | 2 +- test/registered/attention/unittests/dense/test_triton.py | 2 +- .../registered/attention/unittests/dense/test_trtllm_mha.py | 4 ++-- test/registered/attention/unittests/dsa/test_dsa.py | 4 ++-- .../registered/attention/unittests/dsv4/test_deepseek_v4.py | 4 ++-- test/registered/attention/unittests/gdn/test_flashinfer.py | 4 ++-- .../attention/unittests/gdn/test_gdn_replayssm_spec_fold.py | 2 +- .../registered/attention/unittests/gdn/test_torch_native.py | 2 +- test/registered/attention/unittests/gdn/test_triton.py | 2 +- test/registered/attention/unittests/kda/test_triton.py | 2 +- .../registered/attention/unittests/lightning/test_triton.py | 2 +- test/registered/attention/unittests/mamba/test_mamba2.py | 2 +- .../unittests/mamba/test_replay_state_indices_validator.py | 2 +- test/registered/attention/unittests/mla/test_flashinfer.py | 4 ++-- test/registered/attention/unittests/mla/test_flashmla.py | 4 ++-- test/registered/attention/unittests/mla/test_triton.py | 4 ++-- test/registered/attention/unittests/swa/test_flashinfer.py | 4 ++-- .../attention/unittests/swa/test_swa_out_cache_loc.py | 2 +- .../registered/attention/unittests/swa/test_torch_native.py | 2 +- test/registered/attention/unittests/swa/test_triton.py | 2 +- .../registered/backends/test_flashinfer_fusion_preflight.py | 2 +- .../bench_fn/test_bench_serving_reasoning_stream.py | 2 +- test/registered/bench_fn/test_benchmark_datasets_api.py | 2 +- test/registered/bench_fn/test_steady_state_benchmark.py | 2 +- .../chunked_prefill/test_mm_chunked_embedding_unit.py | 2 +- test/registered/chunked_prefill/test_scripted_core_1gpu.py | 2 +- test/registered/chunked_prefill/test_scripted_core_4gpu.py | 2 +- .../constrained_decoding/test_constrained_decoding.py | 2 +- test/registered/core/test_basic_sanity.py | 2 +- test/registered/core/test_basic_sanity_dflash.py | 2 +- test/registered/core/test_basic_sanity_dspark.py | 2 +- test/registered/core/test_basic_sanity_eagle3.py | 2 +- test/registered/core/test_engine_child_pids.py | 2 +- test/registered/core/test_hidden_states.py | 2 +- test/registered/core/test_no_extra_forked_cuda_context.py | 2 +- test/registered/core/test_request_queue_validation.py | 2 +- test/registered/core/test_srt_empty_deps.py | 2 +- test/registered/core/test_srt_endpoint.py | 2 +- test/registered/cp/test_cp_strategy_unit.py | 2 +- test/registered/cp/test_deepseek_v4_flash_fp4_b200_cp.py | 2 +- .../breakable/test_bcg_with_speculative_decoding.py | 2 +- test/registered/dcp/test_dcp_layout_unit.py | 2 +- test/registered/dcp/test_trtllm_mla_family_dcp_metadata.py | 2 +- .../registered/debug_utils/test_tensor_dump_forward_hook.py | 2 +- .../disaggregation/test_disaggregation_aarch64.py | 2 +- test/registered/disaggregation/test_disaggregation_basic.py | 2 +- .../test_disaggregation_chunked_prefill_abort.py | 2 +- .../test_disaggregation_decode_radix_cache_swa.py | 2 +- .../disaggregation/test_disaggregation_hisparse.py | 2 +- .../disaggregation/test_disaggregation_hybrid_attention.py | 2 +- .../disaggregation/test_disaggregation_inkling_mxfp8.py | 2 +- .../disaggregation/test_disaggregation_rust_server.py | 2 +- test/registered/dp_attn/test_dp_attention.py | 2 +- test/registered/dp_engine/test_data_parallelism.py | 2 +- .../test_disaggregation_unified_memory_swa.py | 2 +- .../test_disaggregation_unified_memory_tri.py | 2 +- test/registered/e2e/hicache/test_hicache_unified_memory.py | 2 +- .../e2e/layers/test_logprob_memory_distributed.py | 2 +- .../registered/e2e/lora/test_lora_moe_prefill_cuda_graph.py | 2 +- .../e2e/models/test_deepseek_v4_flash_fp4_megamoe_b200.py | 2 +- test/registered/e2e/models/test_generation_models.py | 2 +- test/registered/e2e/models/test_inkling_small_nvfp4.py | 2 +- test/registered/e2e/models/test_kimi_k3_b300_low_latency.py | 2 +- .../models/test_kimi_linear_unified_memory_dcp_blackwell.py | 2 +- test/registered/e2e/models/test_layernorm_sp.py | 2 +- .../registered/e2e/models/test_transformers_backend_eval.py | 2 +- test/registered/e2e/models/test_transformers_models.py | 2 +- test/registered/e2e/pp/test_pp_spec.py | 2 +- test/registered/e2e/pp/test_pp_spec_embed_scan.py | 2 +- test/registered/e2e/speculative/test_dflash_domino.py | 2 +- test/registered/ep/test_deepep_small_extra.py | 2 +- test/registered/ep/test_tbo_shared_experts_fusion.py | 2 +- test/registered/eplb/test_lplb_distributed.py | 2 +- test/registered/expert_pack/test_expert_pack_runtime.py | 2 +- test/registered/expert_pack/test_kimi_k3_gguf.py | 2 +- test/registered/function_call/test_kimik3_detector.py | 2 +- test/registered/hicache/test_hicache_spec_file_storage.py | 2 +- .../hicache/test_hicache_spec_mooncake_storage.py | 2 +- test/registered/hicache/test_qwen35_hicache.py | 2 +- .../kernels/benchmark/gemm/bench_kda_fp8_skinny_gemm.py | 2 +- .../kernels/ops/attention/test_dcp_lse_combine.py | 2 +- test/registered/kernels/ops/attention/test_dsa_indexer.py | 2 +- .../kernels/ops/attention/test_sm120_paged_mqa_logits.py | 2 +- .../kernels/ops/communication/test_symm_mem_all_gather.py | 2 +- test/registered/kernels/ops/diffusion/test_sites.py | 2 +- .../kernels/ops/embeddings/test_vocab_parallel_embedding.py | 2 +- test/registered/kernels/ops/gemm/test_sm120_fp8_linear.py | 2 +- test/registered/kernels/ops/kv_canary/test_utils.py | 2 +- test/registered/kernels/ops/layernorm/test_mhc_kernels.py | 2 +- .../kernels/ops/speculative/test_boundary_kv_fix_kernels.py | 2 +- test/registered/kv_canary/test_self_e2e_baseline.py | 2 +- test/registered/kv_canary/test_self_e2e_bench_speed.py | 2 +- test/registered/kv_canary/test_self_e2e_pd_baseline.py | 2 +- test/registered/kv_canary/test_self_e2e_pd_perturb.py | 2 +- test/registered/kv_canary/test_self_e2e_perturb_raise.py | 2 +- .../kv_canary/test_self_e2e_perturb_real_kv_unused_cache.py | 2 +- .../kv_canary/test_self_e2e_perturb_real_kv_used.py | 2 +- .../kv_canary/test_self_e2e_perturb_req_to_token.py | 2 +- test/registered/kv_canary/test_self_e2e_pp_baseline.py | 2 +- test/registered/kv_canary/test_self_e2e_pp_perturb.py | 2 +- test/registered/kv_canary/test_self_e2e_pr_25015.py | 2 +- test/registered/kv_canary/test_self_e2e_pr_26329.py | 2 +- test/registered/kv_canary/test_self_unit_buffer_alloc.py | 2 +- test/registered/kv_canary/test_self_unit_capacities.py | 2 +- test/registered/kv_canary/test_self_unit_capture_refusal.py | 2 +- test/registered/kv_canary/test_self_unit_endpoint.py | 2 +- test/registered/kv_canary/test_self_unit_future_tensor.py | 2 +- test/registered/kv_canary/test_self_unit_perturb.py | 2 +- test/registered/kv_canary/test_self_unit_plan_input.py | 2 +- test/registered/kv_canary/test_self_unit_pool_patcher.py | 2 +- .../kv_canary/test_self_unit_pool_patcher_utils.py | 2 +- test/registered/kv_canary/test_self_unit_radix_walker.py | 2 +- .../test_self_unit_req_to_expected_token_ids_manager.py | 2 +- test/registered/kv_canary/test_self_unit_runner_health.py | 2 +- .../kv_canary/test_self_unit_runner_per_forward.py | 2 +- .../kv_canary/test_self_unit_runner_swa_divergence.py | 2 +- test/registered/kv_canary/test_self_unit_runner_sweep.py | 2 +- .../kv_canary/test_self_unit_sweep_plan_builder.py | 2 +- test/registered/kv_canary/test_self_unit_token_oracle.py | 2 +- test/registered/kv_canary/test_self_unit_violation.py | 2 +- test/registered/layers/mamba/test_causal_conv1d.py | 2 +- test/registered/layers/mamba/test_mamba2_mixer.py | 2 +- test/registered/layers/mamba/test_mamba_slot_fused.py | 2 +- test/registered/layers/mamba/test_mamba_ssm.py | 2 +- test/registered/layers/mamba/test_mamba_ssm_ssd.py | 2 +- test/registered/mem_cache/test_int8_checkpoint_store.py | 2 +- test/registered/mem_cache/test_post_capture_kv_sizing.py | 2 +- test/registered/mock_model/test_e2e_cp.py | 2 +- test/registered/mock_model/test_e2e_pd.py | 2 +- test/registered/mock_model/test_e2e_pp.py | 2 +- test/registered/mock_model/test_e2e_spec_eagle.py | 2 +- .../mock_model/test_self_e2e_perturb_next_token_swap.py | 2 +- .../mock_model/test_self_unit_canary_mock_wiring.py | 2 +- test/registered/mock_model/test_self_unit_install.py | 2 +- test/registered/mock_model/test_self_unit_oracle.py | 2 +- .../mock_model/test_self_unit_oracle_torch_vs_ref.py | 2 +- .../mock_model/test_self_unit_sampler_hookpoint.py | 2 +- test/registered/model_loading/test_external_models.py | 2 +- .../model_loading/test_load_weights_from_remote_instance.py | 2 +- test/registered/model_loading/test_startup_weight_load.py | 2 +- test/registered/model_loading/test_weight_cache_daemon.py | 2 +- test/registered/model_loading/test_weight_loader_v2_e2e.py | 2 +- test/registered/moe/test_cutedsl_moe.py | 2 +- test/registered/moe/test_hpc_ops_moe.py | 2 +- test/registered/moe/test_moe_ep_extra.py | 2 +- test/registered/moe/test_zero_experts.py | 2 +- .../observability/test_customized_info_streaming.py | 2 +- .../registered/observability/test_encoder_server_metrics.py | 2 +- test/registered/observability/test_metrics.py | 2 +- test/registered/observability/test_tracing.py | 2 +- .../registered/observability/test_tracing_disaggregation.py | 2 +- .../openai_server/basic/test_openai_completion_rust.py | 2 +- test/registered/openai_server/basic/test_openai_server.py | 2 +- .../openai_server/validation/test_matched_stop.py | 2 +- test/registered/page_major/test_page_major_gpt_oss.py | 2 +- test/registered/page_major/test_page_major_qwen_hybrid.py | 2 +- test/registered/pp/test_pp_parallel_compat.py | 2 +- test/registered/pp/test_pp_single_node.py | 2 +- test/registered/pp/test_pp_single_node_extra.py | 2 +- test/registered/prefill_only/test_embedding_models.py | 2 +- test/registered/prefill_only/test_openai_embedding.py | 2 +- test/registered/prefill_only/test_score_api.py | 2 +- .../radix_cache/test_int8_mamba_checkpoint_e2e.py | 2 +- test/registered/radix_cache/test_mamba2_extra_buffer_kl.py | 2 +- test/registered/radix_cache/test_radix_attention.py | 2 +- test/registered/radix_cache/test_radix_cache_hit.py | 2 +- .../linker/test_unified_cache_linker_kl_dsv4.py | 2 +- .../linker/test_unified_cache_linker_kl_glm52.py | 2 +- .../unified_radix_tree/test_unified_radix_cache_kl_cp.py | 2 +- .../unified_radix_tree/test_unified_radix_cache_kl_dcp.py | 2 +- .../unified_radix_tree/test_unified_radix_cache_kl_dsv4.py | 2 +- .../test_unified_radix_cache_kl_dsv4_pp.py | 2 +- .../unified_radix_tree/test_unified_radix_cache_kl_glm52.py | 2 +- .../test_unified_radix_cache_kl_hybrid_bitexact.py | 2 +- .../unified_radix_tree/test_unified_radix_cache_kl_mamba.py | 2 +- test/registered/rl/test_fp32_lm_head.py | 2 +- .../rl/test_multi_instance_release_memory_occupation.py | 2 +- test/registered/rl/test_patch_torch.py | 2 +- test/registered/rl/test_return_indexer_topk.py | 2 +- test/registered/rl/test_return_routed_experts.py | 2 +- test/registered/rl/test_update_weights_from_distributed.py | 2 +- test/registered/rl/test_update_weights_from_tensor.py | 2 +- test/registered/rotary/test_mrope_axis_map.py | 2 +- test/registered/rotary/test_rope_cache_invalidation.py | 2 +- test/registered/rust/test_rust_extension.py | 2 +- test/registered/sampling/test_deterministic_gumbel_u1.py | 2 +- test/registered/sampling/test_original_logprobs.py | 2 +- test/registered/sampling/test_penalty.py | 2 +- test/registered/scheduler/test_abort_with_metrics.py | 2 +- test/registered/scheduler/test_min_free_slots_delayer.py | 2 +- test/registered/scheduler/test_mixed_chunked_prefill.py | 2 +- test/registered/scheduler/test_priority_scheduling.py | 2 +- test/registered/scheduler/test_retract_decode_logprob.py | 2 +- test/registered/scheduler/test_scheduler_control.py | 2 +- .../scripted_runtime/test_scripted_runtime_core.py | 2 +- test/registered/sessions/test_session_control.py | 2 +- .../spec/dspark/test_dspark_block_accept_estimator.py | 2 +- .../spec/dspark/test_dspark_confidence_metrics.py | 2 +- test/registered/spec/dspark/test_dspark_dp_tier.py | 2 +- .../spec/dspark/test_dspark_draft_path_default.py | 2 +- test/registered/spec/dspark/test_dspark_info_dumper.py | 2 +- test/registered/spec/dspark/test_dspark_kernel_parity.py | 2 +- test/registered/spec/dspark/test_dspark_scheduler.py | 2 +- test/registered/spec/dspark/test_dspark_sps_table.py | 2 +- .../spec/dspark/test_dspark_stacked_ctx_kv_parity.py | 2 +- test/registered/spec/dspark/test_dspark_sts.py | 2 +- test/registered/spec/dspark/test_ragged_verify.py | 2 +- .../spec/dspark/test_ragged_verify_backend_capability.py | 2 +- test/registered/spec/eagle/test_eagle_dp_attention.py | 2 +- test/registered/spec/eagle/test_eagle_reject_sampling.py | 2 +- test/registered/spec/eagle/test_spec_eagle.py | 2 +- test/registered/spec/eagle/test_spec_eagle_fa3.py | 2 +- test/registered/spec/eagle/test_spec_eagle_stress.py | 2 +- test/registered/spec/eagle/test_spec_eagle_triton.py | 2 +- test/registered/spec/test_spec_mixed_chunk.py | 2 +- test/registered/spec/test_spec_ngram.py | 2 +- test/registered/spec/test_spec_standalone.py | 2 +- test/registered/spec/test_spec_standalone_extra.py | 2 +- test/registered/spec/utils/test_build_eagle_tree.py | 2 +- .../batch_overlap/test_tbo_children_dummy_token_mask.py | 2 +- .../batch_overlap/test_tbo_cuda_graph_num_token_device.py | 2 +- .../unit/batch_overlap/test_tbo_filter_batch_marker.py | 2 +- test/registered/unit/beam_search/test_beam_search_core.py | 2 +- test/registered/unit/beam_search/test_fork.py | 2 +- test/registered/unit/beam_search/test_output_decode.py | 2 +- test/registered/unit/bench/test_benchmark_endpoint.py | 2 +- test/registered/unit/cli/test_serve_backends.py | 2 +- .../unit/compilation/test_torch_compile_decoration.py | 2 +- test/registered/unit/configs/test_cohere2_moe_config.py | 2 +- test/registered/unit/configs/test_embedding_model_spec.py | 2 +- test/registered/unit/configs/test_laguna_config.py | 2 +- .../unit/configs/test_linear_attn_model_registry.py | 2 +- test/registered/unit/configs/test_locate_anything_config.py | 2 +- test/registered/unit/configs/test_minicpm_config.py | 2 +- test/registered/unit/configs/test_model_config.py | 2 +- .../unit/configs/test_model_config_parser_registry.py | 2 +- test/registered/unit/configs/test_model_config_scaling.py | 2 +- test/registered/unit/configs/test_model_config_shapes.py | 2 +- .../unit/configs/test_multimodal_piecewise_cuda_graph.py | 2 +- .../registered/unit/configs/test_nano_nemotron_vl_config.py | 2 +- test/registered/unit/configs/test_zaya_config.py | 2 +- .../unit/constrained/test_llguidance_batched_mask.py | 2 +- .../disaggregation/test_admission_abort_not_enqueued.py | 2 +- test/registered/unit/disaggregation/test_dcp_pack.py | 2 +- .../unit/disaggregation/test_decode_hicache_tree_core.py | 2 +- .../unit/disaggregation/test_decode_queue_cleanup.py | 2 +- .../unit/disaggregation/test_decode_req_to_token_pool.py | 2 +- .../unit/disaggregation/test_deferred_decode_kv_release.py | 2 +- .../unit/disaggregation/test_disaggregation_wire.py | 2 +- test/registered/unit/disaggregation/test_encode_receiver.py | 2 +- test/registered/unit/disaggregation/test_encode_server.py | 2 +- test/registered/unit/disaggregation/test_fake_kv_sender.py | 2 +- .../unit/disaggregation/test_kimi_k3_encoder_mode.py | 2 +- test/registered/unit/disaggregation/test_kv_events.py | 2 +- .../unit/disaggregation/test_kv_transfer_replica_metric.py | 2 +- .../test_minimax_sparse_disagg_state_kv_args.py | 2 +- .../unit/disaggregation/test_mooncake_transfer_batching.py | 2 +- .../unit/disaggregation/test_nixl_backend_basic.py | 2 +- .../unit/disaggregation/test_nixl_deferred_kv_release.py | 2 +- .../unit/disaggregation/test_pp_hybrid_kv_transfer.py | 2 +- .../unit/disaggregation/test_pp_mla_kv_transfer.py | 2 +- .../disaggregation/test_prefill_abort_result_cleanup.py | 2 +- .../unit/disaggregation/test_receiver_connection_pool.py | 2 +- .../unit/disaggregation/test_register_to_bootstrap.py | 2 +- .../unit/disaggregation/test_specv2_kvcache_offloading.py | 2 +- .../unit/disaggregation/test_staging_draft_kv_slots.py | 2 +- .../unit/disaggregation/test_unified_memory_move_gate.py | 2 +- test/registered/unit/distributed/test_gated_launch.py | 2 +- .../distributed/test_get_default_distributed_backend.py | 2 +- test/registered/unit/distributed/test_parallel_state.py | 2 +- test/registered/unit/distributed/test_pp_comm_overlap.py | 2 +- test/registered/unit/entrypoints/anthropic/test_serving.py | 2 +- .../unit/entrypoints/openai/test_audio_chunking.py | 2 +- .../unit/entrypoints/openai/test_chat_prompt_round_trip.py | 2 +- test/registered/unit/entrypoints/openai/test_exa_search.py | 2 +- .../registered/unit/entrypoints/openai/test_matched_stop.py | 2 +- test/registered/unit/entrypoints/openai/test_protocol.py | 2 +- .../unit/entrypoints/openai/test_responses_custom_tools.py | 2 +- .../unit/entrypoints/openai/test_responses_protocol.py | 2 +- .../registered/unit/entrypoints/openai/test_serving_chat.py | 2 +- .../unit/entrypoints/openai/test_serving_completions.py | 2 +- .../unit/entrypoints/openai/test_serving_embedding.py | 2 +- .../unit/entrypoints/openai/test_serving_transcription.py | 2 +- .../unit/entrypoints/openai/test_transcription_adapters.py | 2 +- .../unit/entrypoints/openai/test_whisper_adapter.py | 2 +- test/registered/unit/entrypoints/test_grpc_bridge.py | 2 +- test/registered/unit/entrypoints/test_http_server_warmup.py | 2 +- test/registered/unit/entrypoints/test_ready.py | 2 +- .../unit/entrypoints/test_rust_server_dp_ports.py | 2 +- test/registered/unit/entrypoints/test_server_info.py | 2 +- test/registered/unit/entrypoints/test_server_warmup.py | 2 +- test/registered/unit/entrypoints/test_ssl_cert_refresher.py | 2 +- test/registered/unit/entrypoints/test_v1_loads_aggregate.py | 2 +- test/registered/unit/eplb/test_balanced_packing.py | 2 +- .../test_compute_logical_to_rank_dispatch_physical_map.py | 2 +- .../unit/eplb/test_dispatch_dtype_preservation.py | 2 +- test/registered/unit/eplb/test_waterfill_eplb.py | 2 +- .../unit/function_call/test_deepseekv41_detector.py | 2 +- test/registered/unit/function_call/test_dots_detector.py | 2 +- .../unit/function_call/test_function_call_parser.py | 2 +- .../unit/function_call/test_glm47_schema_types.py | 2 +- test/registered/unit/function_call/test_hunyuan_detector.py | 2 +- test/registered/unit/function_call/test_k2_v3_detector.py | 2 +- .../unit/function_call/test_kimik3_structural_tag.py | 2 +- .../unit/function_call/test_minimax_m3_detector.py | 2 +- .../unit/function_call/test_muse_glimmer_format.py | 2 +- test/registered/unit/function_call/test_spark25_detector.py | 2 +- .../hardware_backend/mlx/test_mlx_quantization_override.py | 2 +- .../unit/hardware_backend/mlx/test_mlx_remote_code_gate.py | 2 +- test/registered/unit/kernels/test_fused_op_dispatch.py | 2 +- test/registered/unit/kernels/test_jit_cache.py | 2 +- .../unit/layers/attention/linear/kernels/test_kda_nvidia.py | 2 +- .../unit/layers/attention/linear/kernels/test_kda_ptx.py | 2 +- .../unit/layers/attention/test_dots_hybrid_backend.py | 2 +- .../unit/layers/attention/test_dsa_head_gate_guard.py | 2 +- .../unit/layers/attention/test_dsa_mqa_logits_chunking.py | 2 +- .../layers/attention/test_encoder_decoder_varlen_gather.py | 2 +- .../layers/attention/test_flashattention_graph_metadata.py | 2 +- .../test_flashattention_pa_swa_prefill_lens_size.py | 2 +- .../unit/layers/attention/test_gdn_flashinfer_alignment.py | 2 +- .../layers/attention/test_gdn_prefill_backend_policy.py | 2 +- .../unit/layers/attention/test_index_topk_share.py | 2 +- .../unit/layers/attention/test_kda_helion_dispatcher.py | 2 +- .../registered/unit/layers/attention/test_kv_shard_hooks.py | 2 +- .../unit/layers/attention/test_kv_translate_ownership.py | 2 +- .../unit/layers/attention/test_linear_attn_config.py | 2 +- .../unit/layers/attention/test_mamba_track_state_dtype.py | 2 +- .../unit/layers/attention/test_mla_decode_geometry.py | 2 +- test/registered/unit/layers/attention/test_verify_mask.py | 2 +- .../unit/layers/attention/test_vision_backend_selection.py | 2 +- .../unit/layers/attention/test_vision_max_seqlen.py | 2 +- .../unit/layers/attention/test_vision_strided_qkv.py | 2 +- .../unit/layers/moe/test_deepep_v2_buffer_lifecycle.py | 2 +- .../unit/layers/moe/test_deepep_v2_masked_slab.py | 2 +- .../registered/unit/layers/moe/test_deepep_v2_wire_dtype.py | 2 +- .../unit/layers/moe/test_flashinfer_dispatcher.py | 2 +- test/registered/unit/layers/moe/test_flashinfer_megamoe.py | 2 +- test/registered/unit/layers/moe/test_fused_moe_native.py | 2 +- .../unit/layers/moe/test_fused_shared_expert_scaling.py | 2 +- .../unit/layers/moe/test_mega_moe_deepgemm_api.py | 2 +- .../unit/layers/moe/test_topk_correction_bias_cache.py | 2 +- test/registered/unit/layers/moe/test_w4afp8_deepep_dtype.py | 2 +- .../unit/layers/moe/test_w4afp8_deepep_post_reorder.py | 2 +- .../layers/quantization/test_compressed_tensors_kv_cache.py | 2 +- .../layers/quantization/test_compressed_tensors_lm_head.py | 2 +- .../quantization/test_compressed_tensors_mixed_precision.py | 2 +- .../test_compressed_tensors_wna16_moe_no_linear.py | 2 +- .../unit/layers/quantization/test_deepgemm_ue8m0_requant.py | 2 +- .../quantization/test_flashinfer_trtllm_fp8_fallback.py | 2 +- .../layers/quantization/test_fp4_kv_cache_quant_method.py | 2 +- .../quantization/test_fp8_blockwise_linear_backends.py | 6 +++--- .../layers/quantization/test_fp8_moe_runner_fallback.py | 2 +- .../layers/quantization/test_fp8_moe_runner_ownership.py | 2 +- .../unit/layers/quantization/test_fp8_moe_weight_gating.py | 2 +- .../unit/layers/quantization/test_humming_w4afp8_schemas.py | 2 +- .../unit/layers/quantization/test_int8_linear_methods.py | 2 +- .../unit/layers/quantization/test_modelopt_nvfp4.py | 2 +- .../layers/quantization/test_modelopt_nvfp4_moe_dispatch.py | 2 +- .../layers/quantization/test_modelopt_nvfp4_moe_scales.py | 2 +- .../quantization/test_mxfp4_flashinfer_activation_prep.py | 2 +- .../unit/layers/quantization/test_mxfp4_sm100_trtllm_gen.py | 2 +- .../unit/layers/quantization/test_mxfp4_sm120_cutlass.py | 2 +- .../unit/layers/quantization/test_mxfp4_sm90_cutlass.py | 2 +- .../unit/layers/quantization/test_nvfp4_linear_backends.py | 2 +- .../unit/layers/quantization/test_nvfp4_moe_backends.py | 2 +- .../unit/layers/quantization/test_quark_config.py | 2 +- .../registered/unit/layers/quantization/test_quark_utils.py | 2 +- .../layers/quantization/test_unquant_apply_with_addend.py | 2 +- test/registered/unit/layers/test_bailing_mrope_shape.py | 2 +- test/registered/unit/layers/test_conv_layer.py | 2 +- test/registered/unit/layers/test_dsv41_candidate_blocks.py | 2 +- .../registered/unit/layers/test_dsv4_kv_splits_heuristic.py | 2 +- test/registered/unit/layers/test_dsv4_nonpaged_indexer.py | 2 +- .../registered/unit/layers/test_flashattention_paged_mha.py | 2 +- test/registered/unit/layers/test_flashinfer_comm_fusion.py | 2 +- test/registered/unit/layers/test_fp8_bpreshuffle_scale.py | 2 +- test/registered/unit/layers/test_gdn_mis_metadata.py | 2 +- .../registered/unit/layers/test_kpool_planner_cpu_mirror.py | 2 +- test/registered/unit/layers/test_kpool_stream_scheduling.py | 2 +- test/registered/unit/layers/test_logprob_chunk_stitching.py | 2 +- test/registered/unit/layers/test_logprob_fast_input.py | 2 +- .../registered/unit/layers/test_mamba2_track_ssm_indices.py | 2 +- .../unit/layers/test_mamba_prefill_track_metadata.py | 2 +- .../unit/layers/test_mamba_state_scatter_triton.py | 4 ++-- .../unit/layers/test_minicpm_attention_adapter.py | 2 +- test/registered/unit/layers/test_minicpm_sparse_cache.py | 2 +- test/registered/unit/layers/test_minicpm_sparse_metadata.py | 2 +- test/registered/unit/layers/test_mova.py | 2 +- test/registered/unit/layers/test_pooler_score_and_pool.py | 2 +- test/registered/unit/layers/test_radix_attention.py | 2 +- test/registered/unit/layers/test_radix_linear_attention.py | 2 +- test/registered/unit/layers/test_yarn_cache_extension.py | 2 +- test/registered/unit/lora/test_eviction_policy.py | 2 +- .../unit/lora/test_experimental_sgl_marlin_runtime_unit.py | 2 +- .../unit/lora/test_inkling_linearized_lora_unit.py | 2 +- test/registered/unit/lora/test_lm_head_pruning.py | 2 +- test/registered/unit/lora/test_lora_moe_inplace_unit.py | 2 +- .../unit/lora/test_lora_spec_verify_batch_info.py | 2 +- test/registered/unit/lora/test_mem_pool_ep_unit.py | 2 +- .../unit/managers/scheduler_components/test_dp_attn.py | 2 +- .../managers/scheduler_components/test_invariant_checker.py | 2 +- .../managers/scheduler_components/test_output_sender.py | 2 +- test/registered/unit/managers/test_auxiliary_output.py | 2 +- .../unit/managers/test_bailing_modality_metadata.py | 2 +- .../managers/test_batch_result_processor_hidden_states.py | 2 +- .../managers/test_batch_result_processor_mamba_boundary.py | 2 +- .../managers/test_batch_result_processor_spec_grammar.py | 2 +- .../unit/managers/test_data_parallel_controller.py | 2 +- test/registered/unit/managers/test_detailed_annotations.py | 2 +- .../unit/managers/test_disagg_idle_step_counters.py | 2 +- test/registered/unit/managers/test_embed_overrides.py | 2 +- test/registered/unit/managers/test_fanout_communicator.py | 2 +- .../unit/managers/test_finish_length_speculative.py | 2 +- test/registered/unit/managers/test_flat_raw_top_logprobs.py | 2 +- .../unit/managers/test_grammar_stop_speculative.py | 2 +- .../unit/managers/test_hidden_state_server_mode.py | 2 +- test/registered/unit/managers/test_hisparse_unit.py | 2 +- test/registered/unit/managers/test_kv_page_invariants.py | 2 +- test/registered/unit/managers/test_load_inquirer.py | 2 +- .../registered/unit/managers/test_load_snapshot_backends.py | 2 +- test/registered/unit/managers/test_loadstat_wire.py | 2 +- .../unit/managers/test_lora_update_result_merge.py | 2 +- .../registered/unit/managers/test_mamba_checkpoint_depth.py | 2 +- test/registered/unit/managers/test_mm_hashes.py | 2 +- test/registered/unit/managers/test_mm_process_config.py | 2 +- .../registered/unit/managers/test_mm_shm_error_consensus.py | 2 +- test/registered/unit/managers/test_mm_utils_split.py | 2 +- test/registered/unit/managers/test_multi_tokenizer_mixin.py | 2 +- .../unit/managers/test_multimodal_abort_cleanup.py | 2 +- .../unit/managers/test_output_streamer_customized_info.py | 2 +- .../unit/managers/test_output_streamer_logprobs.py | 2 +- test/registered/unit/managers/test_pp_cp_rank_offsets.py | 2 +- test/registered/unit/managers/test_prefill_adder.py | 2 +- .../unit/managers/test_prefill_delayer_high_watermark.py | 2 +- .../managers/test_priority_scheduling_disaggregation.py | 2 +- test/registered/unit/managers/test_retraction_order.py | 2 +- .../unit/managers/test_sampling_mask_validation.py | 2 +- .../test_schedule_batch_convert_decode_to_extend.py | 2 +- .../unit/managers/test_schedule_batch_out_of_place.py | 2 +- .../unit/managers/test_schedule_batch_prepare_for_decode.py | 2 +- .../unit/managers/test_schedule_batch_req_pool_indices.py | 2 +- test/registered/unit/managers/test_schedule_policy.py | 2 +- .../unit/managers/test_scheduler_chunked_abort_race.py | 2 +- .../unit/managers/test_scheduler_chunked_req_gate.py | 2 +- test/registered/unit/managers/test_scheduler_flush_cache.py | 2 +- .../unit/managers/test_scheduler_hicache_attach.py | 2 +- .../unit/managers/test_scheduler_hicache_events.py | 2 +- .../unit/managers/test_scheduler_init_req_max_new_tokens.py | 2 +- .../registered/unit/managers/test_scheduler_on_idle_load.py | 2 +- .../unit/managers/test_scheduler_pause_generation.py | 2 +- .../unit/managers/test_scheduler_prefill_decode_interval.py | 2 +- .../registered/unit/managers/test_scheduler_recv_skipper.py | 2 +- .../managers/test_scheduler_sampling_mask_validation.py | 2 +- test/registered/unit/managers/test_scheduler_timeouts.py | 2 +- .../unit/managers/test_scheduler_weight_version_tracking.py | 2 +- test/registered/unit/managers/test_stop_str_speculative.py | 2 +- .../unit/managers/test_tokenizer_config_updates.py | 2 +- .../unit/managers/test_tokenizer_manager_rid_cleanup.py | 2 +- test/registered/unit/managers/test_trim_matched_stop.py | 2 +- test/registered/unit/managers/test_vocab_boundary_finish.py | 2 +- test/registered/unit/mem_cache/test_asymmetric_mha_pool.py | 2 +- .../unit/mem_cache/test_asymmetric_mha_pool_host_unit.py | 2 +- .../registered/unit/mem_cache/test_decode_radix_lock_ref.py | 2 +- .../unit/mem_cache/test_decode_retraction_backup.py | 2 +- test/registered/unit/mem_cache/test_dllm_fdfo_kv_reuse.py | 2 +- .../registered/unit/mem_cache/test_dsa_layer_shard_utils.py | 2 +- .../unit/mem_cache/test_dsa_layer_split_broadcast.py | 2 +- test/registered/unit/mem_cache/test_dsa_pool_host_unit.py | 2 +- .../unit/mem_cache/test_dsv4_c4_state_lifecycle.py | 2 +- .../unit/mem_cache/test_dsv4_compress_write_pad.py | 2 +- .../registered/unit/mem_cache/test_dsv4_compressed_pools.py | 2 +- test/registered/unit/mem_cache/test_dsv4_hicache_l2.py | 2 +- .../registered/unit/mem_cache/test_dsv4_unified_fp8_pool.py | 2 +- .../unit/mem_cache/test_embedding_cache_controller.py | 2 +- test/registered/unit/mem_cache/test_evict_policy.py | 2 +- .../unit/mem_cache/test_flashkda_strided_state_access.py | 2 +- test/registered/unit/mem_cache/test_free_kv_row_coalesce.py | 2 +- test/registered/unit/mem_cache/test_full_loc_fast_path.py | 2 +- test/registered/unit/mem_cache/test_hicache_auto_size.py | 2 +- test/registered/unit/mem_cache/test_hicache_copy_rounds.py | 2 +- .../registered/unit/mem_cache/test_hicache_dcp_host_pool.py | 2 +- .../registered/unit/mem_cache/test_hicache_file_lru_unit.py | 2 +- .../registered/unit/mem_cache/test_hicache_host_register.py | 2 +- .../unit/mem_cache/test_hicache_load_back_timing.py | 2 +- test/registered/unit/mem_cache/test_hicache_nixl_cleaner.py | 2 +- test/registered/unit/mem_cache/test_hicache_nixl_storage.py | 2 +- .../mem_cache/test_hicache_staged_write_back_dispatch.py | 2 +- test/registered/unit/mem_cache/test_hisparse_allocator.py | 2 +- .../unit/mem_cache/test_hisparse_max_token_pool_size.py | 2 +- .../unit/mem_cache/test_hisparse_slot_translation.py | 2 +- test/registered/unit/mem_cache/test_host_memory.py | 2 +- .../registered/unit/mem_cache/test_hybrid_pool_assembler.py | 2 +- .../unit/mem_cache/test_inkling_sconv_strided_conv_state.py | 2 +- .../unit/mem_cache/test_kda_fused_decode_strided_state.py | 2 +- .../registered/unit/mem_cache/test_linker_pool_assembler.py | 2 +- .../unit/mem_cache/test_mamba_donated_alloc_ratio.py | 2 +- test/registered/unit/mem_cache/test_mamba_path_state_cap.py | 4 ++-- .../unit/mem_cache/test_mamba_state_transfer_buffers.py | 2 +- test/registered/unit/mem_cache/test_mamba_unittest.py | 2 +- test/registered/unit/mem_cache/test_mem_cache_utils.py | 2 +- test/registered/unit/mem_cache/test_mem_pool_host.py | 2 +- .../unit/mem_cache/test_minimax_sparse_pool_host_unit.py | 2 +- .../unit/mem_cache/test_minimax_sparse_pool_pd_unit.py | 2 +- .../unit/mem_cache/test_mla_host_dedup_primitives.py | 2 +- test/registered/unit/mem_cache/test_mmap_allocator.py | 2 +- .../unit/mem_cache/test_mooncake_group_semantics.py | 2 +- .../unit/mem_cache/test_mooncake_ssd_offload_path.py | 2 +- .../unit/mem_cache/test_mooncake_standalone_dummy_mamba.py | 2 +- .../registered/unit/mem_cache/test_mooncake_store_config.py | 2 +- .../unit/mem_cache/test_mooncake_tenant_config.py | 2 +- .../registered/unit/mem_cache/test_multi_ended_allocator.py | 2 +- .../unit/mem_cache/test_mxfp8_mha_pool_host_unit.py | 2 +- .../unit/mem_cache/test_mxfp8_scale_transfer_buffers.py | 2 +- .../unit/mem_cache/test_npu_mamba_async_layout.py | 2 +- .../registered/unit/mem_cache/test_page_interleave_shard.py | 2 +- .../unit/mem_cache/test_paged_allocator_lazy_release.py | 2 +- test/registered/unit/mem_cache/test_paged_free_segment.py | 2 +- .../unit/mem_cache/test_pd_envelope_transfer_layout.py | 2 +- test/registered/unit/mem_cache/test_pp_prefetch_ticket.py | 2 +- .../registered/unit/mem_cache/test_prefill_memory_budget.py | 2 +- test/registered/unit/mem_cache/test_pure_swa_radix_cache.py | 2 +- test/registered/unit/mem_cache/test_quantized_kv_pool.py | 2 +- test/registered/unit/mem_cache/test_radix_cache_unit.py | 2 +- test/registered/unit/mem_cache/test_radix_force_miss.py | 2 +- test/registered/unit/mem_cache/test_registry.py | 2 +- .../unit/mem_cache/test_replayssm_ring_accounting.py | 2 +- .../unit/mem_cache/test_retraction_mamba_backup.py | 2 +- .../unit/mem_cache/test_rust_tree_core_integration.py | 2 +- .../unit/mem_cache/test_rust_unified_radix_cache_bench.py | 2 +- .../unit/mem_cache/test_session_token_share_unit.py | 2 +- .../unit/mem_cache/test_session_unified_radix_cache.py | 2 +- .../unit/mem_cache/test_storage_prefetch_lifecycle.py | 2 +- .../unit/mem_cache/test_streaming_session_unit.py | 2 +- .../unit/mem_cache/test_swa_alloc_extend_page_estimation.py | 2 +- .../unit/mem_cache/test_swa_locked_full_recover_unified.py | 2 +- test/registered/unit/mem_cache/test_swa_pool_v_head_dim.py | 2 +- test/registered/unit/mem_cache/test_swa_ring_page_return.py | 2 +- test/registered/unit/mem_cache/test_tlru_eviction_policy.py | 2 +- test/registered/unit/mem_cache/test_tree_core_registry.py | 2 +- test/registered/unit/mem_cache/test_umbp_host_allocator.py | 2 +- .../unit/mem_cache/test_unified_byte_accounting.py | 2 +- test/registered/unit/mem_cache/test_unified_cache_linker.py | 4 ++-- .../registered/unit/mem_cache/test_unified_capacity_memo.py | 2 +- .../unit/mem_cache/test_unified_free_no_host_sync.py | 2 +- .../unit/mem_cache/test_unified_handout_zeroing.py | 2 +- .../unit/mem_cache/test_unified_hicache_regressions.py | 2 +- test/registered/unit/mem_cache/test_unified_mamba_views.py | 2 +- test/registered/unit/mem_cache/test_unified_mha_views.py | 2 +- .../unit/mem_cache/test_unified_mla_block_table.py | 2 +- .../unit/mem_cache/test_unified_mla_gpu_parity.py | 2 +- test/registered/unit/mem_cache/test_unified_mla_views.py | 2 +- test/registered/unit/mem_cache/test_unified_npool_sweep.py | 2 +- .../mem_cache/test_unified_radix_allocation_eviction.py | 2 +- .../unit/mem_cache/test_unified_radix_hicache_dispatch.py | 2 +- test/registered/unit/mem_cache/test_unified_tri_pool.py | 2 +- test/registered/unit/memory/test_ptr_table.py | 2 +- .../model_runner_components/test_attention_backend_setup.py | 2 +- .../model_runner_components/test_cuda_graph_setup.py | 2 +- .../model_runner_components/test_layer_setup.py | 2 +- .../model_runner_components/test_ngram_embedding_manager.py | 2 +- .../model_runner_components/test_spec_aux_hidden_state.py | 2 +- .../model_runner_components/test_startup_weight_load.py | 2 +- .../model_executor/runner/test_decode_cuda_graph_runner.py | 2 +- .../runner/test_decode_cuda_graph_shared_read_fence.py | 2 +- .../model_executor/runner/test_flashinfer_autotune_sync.py | 2 +- .../runner/test_hidden_state_graph_recapture.py | 2 +- .../runner/test_prefill_cuda_graph_padding.py | 2 +- .../model_executor/runner/test_prefill_shared_read_done.py | 2 +- .../runner_backend/test_backend_resolution.py | 2 +- .../runner_backend/test_full_cuda_graph_backend.py | 2 +- .../model_executor/runner_utils/test_graph_pool_borrow.py | 2 +- .../unit/model_executor/test_chunked_prefix_cache_gate.py | 2 +- .../model_executor/test_compensated_mhc_update_guard.py | 2 +- .../unit/model_executor/test_cuda_graph_buffer_registry.py | 2 +- .../model_executor/test_forward_metadata_plan_record.py | 2 +- .../unit/model_executor/test_hisparse_pool_configurator.py | 2 +- .../unit/model_executor/test_kv_canary_headroom.py | 2 +- .../unit/model_executor/test_mlp_sync_pad_unpad.py | 2 +- .../unit/model_executor/test_model_runner_decode_rows.py | 2 +- .../test_num_token_non_padded_localization.py | 2 +- .../unit/model_executor/test_pool_configurator.py | 2 +- .../unit/model_executor/test_prefill_cuda_graph_runner.py | 2 +- .../test_prefill_cuda_graph_runner_helpers.py | 2 +- .../model_executor/test_unified_out_cache_loc_rebind.py | 2 +- .../unit/model_loader/test_default_model_loader.py | 2 +- test/registered/unit/model_loader/test_modelopt_loader.py | 2 +- .../unit/model_loader/test_prefetch_checkpoints.py | 2 +- test/registered/unit/model_loader/test_presharded_loader.py | 2 +- .../unit/model_loader/test_remote_instance_loader.py | 2 +- .../unit/model_loader/test_runai_model_streamer_loader.py | 2 +- .../unit/model_loader/test_transformers_fallback.py | 2 +- .../unit/model_loader/test_weight_cache_protocol.py | 2 +- test/registered/unit/model_loader/test_weight_utils.py | 2 +- test/registered/unit/models/test_bailing_vl_loader.py | 2 +- test/registered/unit/models/test_cosmos3.py | 2 +- test/registered/unit/models/test_cosmos3_edge.py | 2 +- test/registered/unit/models/test_deepseek_mla_dispatch.py | 2 +- test/registered/unit/models/test_deepseek_nextn_mm_embed.py | 2 +- .../unit/models/test_deepseek_pp_mtp_embedding.py | 2 +- .../unit/models/test_deepseek_v4_amd_fused_mhc.py | 2 +- .../models/test_deepseek_v4_mxfp4_shared_expert_requant.py | 2 +- test/registered/unit/models/test_deepseek_v4_rope_policy.py | 2 +- .../unit/models/test_deepseek_v4_shared_expert_fusion.py | 2 +- test/registered/unit/models/test_draft_entry_hook_parity.py | 2 +- test/registered/unit/models/test_glm5_next_bfg_fusion.py | 2 +- test/registered/unit/models/test_glm5_next_modelopt.py | 2 +- test/registered/unit/models/test_glm_moe_gate_fp32.py | 2 +- test/registered/unit/models/test_glm_nextn_moe_ptpc.py | 2 +- .../unit/models/test_glmmoedsa_correction_bias_fp32.py | 2 +- test/registered/unit/models/test_gpt_oss_runai_ownership.py | 2 +- .../unit/models/test_granitemoe_split_expert_loading.py | 2 +- .../unit/models/test_hunyuan_v3_nextn_weight_loading.py | 2 +- test/registered/unit/models/test_hunyuan_v4.py | 2 +- .../unit/models/test_hunyuan_v4_nextn_weight_loading.py | 2 +- test/registered/unit/models/test_interns1pro_processor.py | 2 +- test/registered/unit/models/test_kimi_k25.py | 2 +- test/registered/unit/models/test_kimi_k3_vision.py | 2 +- test/registered/unit/models/test_kimi_vl.py | 2 +- test/registered/unit/models/test_kimi_vl_moonvit.py | 2 +- test/registered/unit/models/test_llava.py | 2 +- test/registered/unit/models/test_llava_processor_pool.py | 2 +- test/registered/unit/models/test_locate_anything.py | 2 +- .../unit/models/test_longcat_flash_router_hpc_gemm.py | 2 +- test/registered/unit/models/test_mellum.py | 2 +- test/registered/unit/models/test_mistral3_vision_feature.py | 2 +- test/registered/unit/models/test_moss_vl_processor.py | 2 +- test/registered/unit/models/test_nano_nemotron_vl.py | 2 +- test/registered/unit/models/test_nemotron_h_mtp.py | 2 +- test/registered/unit/models/test_nemotron_h_shared_add.py | 2 +- .../unit/models/test_nemotron_h_weight_loading.py | 2 +- test/registered/unit/models/test_paddleocr_vl_vision.py | 2 +- test/registered/unit/models/test_qwen3_5_modelopt_fp4.py | 2 +- .../unit/models/test_qwen3_5_packed_weight_loader.py | 2 +- .../unit/models/test_qwen3_5_pipeline_parallel.py | 2 +- .../unit/models/test_qwen3_embedding_registration.py | 2 +- .../unit/models/test_qwen3_vl_feature_materialization.py | 2 +- test/registered/unit/models/test_qwen4_exp_ple_table.py | 2 +- test/registered/unit/models/test_radio.py | 2 +- test/registered/unit/models/test_rocm_hisparse_fused_kv.py | 2 +- .../unit/models/test_shared_experts_fusion_gates.py | 2 +- test/registered/unit/models/test_xllm.py | 2 +- test/registered/unit/models/test_zaya_cca.py | 2 +- test/registered/unit/models/test_zaya_mod_tp.py | 2 +- .../unit/multimodal/rust/inkling/test_bindings.py | 2 +- test/registered/unit/multimodal/rust/qwen/test_driver.py | 2 +- .../registered/unit/multimodal/rust/qwen/test_e2e_parity.py | 2 +- .../registered/unit/multimodal/rust/qwen/test_preprocess.py | 2 +- .../unit/multimodal/rust/qwen/test_rust_mm_processor.py | 2 +- .../unit/multimodal/rust/qwen/test_token_layout_mrope.py | 2 +- test/registered/unit/multimodal/rust/shared/test_fetch.py | 2 +- .../unit/multimodal/rust/shared/test_image_decode.py | 2 +- .../unit/multimodal/rust/shared/test_partition_cores.py | 2 +- .../unit/multimodal/rust/shared/test_rust_mm_gate.py | 2 +- .../multimodal/rust/shared/test_rust_server_extension.py | 2 +- .../unit/multimodal/rust/shared/test_wrap_encoded.py | 2 +- .../unit/multimodal/test_audio_container_decode.py | 2 +- .../registered/unit/multimodal/test_bailing_mm_processor.py | 2 +- .../unit/multimodal/test_base_processor_bad_input.py | 2 +- .../unit/multimodal/test_base_processor_image_decode.py | 2 +- .../registered/unit/multimodal/test_cuda_ipc_pool_budget.py | 2 +- test/registered/unit/multimodal/test_cuda_vmm_transport.py | 2 +- .../unit/multimodal/test_deepseek_ocr_geometry.py | 2 +- test/registered/unit/multimodal/test_dots_note_omni.py | 2 +- test/registered/unit/multimodal/test_evs.py | 2 +- .../unit/multimodal/test_feature_materialization.py | 2 +- test/registered/unit/multimodal/test_glm4v_mixed_offsets.py | 2 +- .../unit/multimodal/test_gpu_feature_transport.py | 2 +- .../unit/multimodal/test_kimi_k3_gpu_preprocess.py | 2 +- .../unit/multimodal/test_media_artifact_processor.py | 2 +- test/registered/unit/multimodal/test_media_url_security.py | 2 +- test/registered/unit/multimodal/test_mrope_encoder_utils.py | 2 +- .../unit/multimodal/test_nano_nemotron_vl_processor.py | 2 +- test/registered/unit/multimodal/test_pixtral_processor.py | 2 +- .../multimodal/test_precomputed_embedding_validation.py | 2 +- test/registered/unit/multimodal/test_preprocess_cache.py | 2 +- .../unit/multimodal/test_processor_async_call_sites.py | 2 +- .../unit/multimodal/test_processor_clone_isolation.py | 2 +- .../unit/multimodal/test_processor_device_selection.py | 2 +- .../unit/multimodal/test_tensor_transport_mode.py | 2 +- .../unit/multimodal/test_vit_cuda_graph_runner.py | 2 +- .../registered/unit/multimodal/test_vit_npu_graph_runner.py | 2 +- .../unit/npu/test_sparsity_driven_kv_offload_config.py | 2 +- test/registered/unit/observability/test_cpu_monitor.py | 2 +- .../unit/observability/test_forward_pass_metrics.py | 2 +- test/registered/unit/observability/test_func_timer.py | 2 +- test/registered/unit/observability/test_label_transform.py | 2 +- test/registered/unit/observability/test_metrics_utils.py | 2 +- test/registered/unit/observability/test_ray_wrappers.py | 2 +- test/registered/unit/observability/test_req_time_stats.py | 2 +- .../unit/observability/test_request_metrics_exporter.py | 2 +- .../unit/observability/test_scheduler_stage_metrics.py | 2 +- .../unit/observability/test_startup_func_log_and_timer.py | 2 +- test/registered/unit/observability/test_stat_loggers_di.py | 2 +- test/registered/unit/observability/test_trace.py | 2 +- test/registered/unit/parser/test_code_completion_parser.py | 2 +- test/registered/unit/parser/test_conversation.py | 2 +- test/registered/unit/parser/test_harmony_parser.py | 2 +- test/registered/unit/parser/test_hunyuan_reasoning.py | 2 +- test/registered/unit/parser/test_inkling_renderer.py | 2 +- test/registered/unit/parser/test_jinja_template_utils.py | 2 +- test/registered/unit/parser/test_k2_v3_reasoning_parser.py | 2 +- test/registered/unit/parser/test_kimik3_reasoning_parser.py | 2 +- test/registered/unit/parser/test_reasoning_parser.py | 2 +- test/registered/unit/platforms/test_platform_interface.py | 2 +- test/registered/unit/plugins/test_hook_registry.py | 2 +- test/registered/unit/plugins/test_load_plugins.py | 2 +- test/registered/unit/sampling/test_ascend_sampler.py | 2 +- .../registered/unit/sampling/test_custom_logit_processor.py | 2 +- test/registered/unit/sampling/test_penaltylib.py | 2 +- test/registered/unit/sampling/test_sampling_batch_info.py | 2 +- .../unit/sampling/test_sampling_metadata_staging.py | 2 +- test/registered/unit/sampling/test_sampling_params.py | 2 +- test/registered/unit/server_args/test_declared_fallbacks.py | 2 +- test/registered/unit/server_args/test_model_config_cache.py | 2 +- test/registered/unit/server_args/test_model_source_paths.py | 2 +- .../unit/server_args/test_page_major_backend_allowlist.py | 2 +- .../server_args/test_platform_prefill_cp_deprecation.py | 2 +- .../unit/server_args/test_record_holds_the_raw_input.py | 2 +- .../unit/server_args/test_resolution_declarations.py | 2 +- .../unit/server_args/test_resolution_hook_registry.py | 2 +- .../unit/server_args/test_resolution_is_reproducible.py | 4 ++-- test/registered/unit/server_args/test_server_args.py | 2 +- .../server_args/test_unified_prefill_cuda_graph_gate.py | 2 +- test/registered/unit/server_args/test_unified_tbo_gate.py | 2 +- test/registered/unit/spec/test_adaptive_runtime_state.py | 2 +- test/registered/unit/spec/test_adaptive_spec_params.py | 2 +- .../unit/spec/test_decode_bookkeeping_ownership.py | 2 +- test/registered/unit/spec/test_dflash_extra_buffer_lazy.py | 2 +- test/registered/unit/spec/test_dflash_logits.py | 2 +- test/registered/unit/spec/test_dflash_overlap_hostsync.py | 2 +- .../unit/spec/test_draft_construction_isolation.py | 2 +- test/registered/unit/spec/test_draft_per_runner_config.py | 2 +- .../unit/spec/test_dspark_target_hidden_projection.py | 2 +- .../unit/spec/test_eagle_draft_cuda_graph_runner.py | 2 +- test/registered/unit/spec/test_eagle_draft_extend_logits.py | 2 +- test/registered/unit/spec/test_eagle_gate_routing.py | 2 +- test/registered/unit/spec/test_eagle_seeded_coins.py | 2 +- .../unit/spec/test_eagle_worker_v2_topk1_fastpath.py | 2 +- test/registered/unit/spec/test_fast_prefill_plan.py | 2 +- test/registered/unit/spec/test_ngram_corpus.py | 2 +- test/registered/unit/spec/test_ngram_mamba_verify_update.py | 2 +- test/registered/unit/spec/test_oot_dflash_hooks.py | 2 +- test/registered/unit/spec/test_plugin_hook_signatures.py | 2 +- .../unit/spec/test_spec_cpu_overlap_constraint.py | 2 +- test/registered/unit/spec/test_spec_utils_traverse_tree.py | 2 +- test/registered/unit/spec/test_uno_tree_config.py | 2 +- test/registered/unit/spec/test_uno_tree_sparse_sampling.py | 2 +- .../state_capturer/test_routed_experts_scattered_a2a.py | 2 +- test/registered/unit/test_checkpoint_quantization.py | 2 +- test/registered/unit/test_cuda_vmm_utils.py | 2 +- test/registered/unit/test_dsa_tilelang_fp8_validation.py | 2 +- test/registered/unit/test_environ.py | 2 +- test/registered/unit/test_flashinfer_sparse_mla.py | 2 +- test/registered/unit/test_global_config_read_ratchet.py | 2 +- test/registered/unit/test_model_overrides.py | 2 +- test/registered/unit/test_platform_context.py | 2 +- test/registered/unit/test_pre_publish_readers.py | 2 +- test/registered/unit/test_quantization_post_load.py | 2 +- test/registered/unit/test_rocm_pageable_h2d_staging.py | 2 +- test/registered/unit/test_runai_utils.py | 2 +- test/registered/unit/test_runtime_context.py | 2 +- test/registered/unit/test_runtime_context_config_bags.py | 2 +- test/registered/unit/test_runtime_context_override.py | 2 +- test/registered/unit/test_server_args_cli_metadata.py | 2 +- test/registered/unit/test_server_args_namespaces.py | 2 +- .../unit/test_split_attention_backend_decisions.py | 2 +- .../unit/tokenizer/test_mistral_empty_assistant.py | 2 +- .../unit/tokenizer/test_tekken_tokenizer_routing.py | 2 +- test/registered/unit/tokenizer/test_tiktoken_tokenizer.py | 2 +- test/registered/unit/utils/test_common.py | 2 +- test/registered/unit/utils/test_diffusion_torch_fallback.py | 2 +- test/registered/unit/utils/test_field_validators.py | 2 +- test/registered/unit/utils/test_gauge_histogram.py | 2 +- test/registered/unit/utils/test_hf_transformers.py | 2 +- .../registered/unit/utils/test_hf_transformers_fastokens.py | 2 +- test/registered/unit/utils/test_hf_transformers_loading.py | 2 +- test/registered/unit/utils/test_invariants.py | 2 +- test/registered/unit/utils/test_json_response.py | 2 +- test/registered/unit/utils/test_profile_merger.py | 2 +- test/registered/unit/utils/test_subprocess_watchdog.py | 2 +- test/registered/unit/utils/test_weight_versions.py | 2 +- test/registered/utils/test_log_utils.py | 2 +- test/registered/utils/test_network_address.py | 2 +- test/registered/utils/test_numa_utils.py | 6 +++--- test/registered/utils/test_phase_checker.py | 2 +- test/registered/utils/test_socket_utils.py | 2 +- test/registered/utils/test_stale_shm_cleanup.py | 2 +- test/registered/vlm/test_video_utils.py | 2 +- test/registered/vlm/test_vision_openai_server_extra.py | 2 +- test/registered/xpu/test_numa_utils_xpu.py | 2 +- 805 files changed, 825 insertions(+), 825 deletions(-) diff --git a/test/registered/attention/test_aiter_gluon_h12_fp8.py b/test/registered/attention/test_aiter_gluon_h12_fp8.py index aa759cb3b3..afc4a314c9 100644 --- a/test/registered/attention/test_aiter_gluon_h12_fp8.py +++ b/test/registered/attention/test_aiter_gluon_h12_fp8.py @@ -15,7 +15,7 @@ from sglang.srt.layers.attention import aiter_mla_gluon as mod from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=15, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _GLUON_FN = "sglang.srt.layers.attention.aiter_mla_gluon._gluon_fn" diff --git a/test/registered/attention/test_create_kvindices.py b/test/registered/attention/test_create_kvindices.py index 5124229d8f..f17cf52d93 100644 --- a/test/registered/attention/test_create_kvindices.py +++ b/test/registered/attention/test_create_kvindices.py @@ -9,7 +9,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase # Triton kernel unit test for KV indices creation -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/attention/test_hybrid_attn_backend.py b/test/registered/attention/test_hybrid_attn_backend.py index d6564d5c04..df93fb6585 100644 --- a/test/registered/attention/test_hybrid_attn_backend.py +++ b/test/registered/attention/test_hybrid_attn_backend.py @@ -11,7 +11,7 @@ from sglang.test.test_utils import ( # Hybrid attention backend tests (FA3 prefill + FlashInfer decode, requires SM 90+ / H100) # Multiple test classes: base, MLA, TorchCompile, SpecDecode variants -register_cuda_ci(est_time=393, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=520, stage="extra-a", runner_config="1-gpu-large") class TestHybridAttnBackendMLA(TestHybridAttnBackendBase): diff --git a/test/registered/attention/test_qwen35_deterministic.py b/test/registered/attention/test_qwen35_deterministic.py index 60b9fee3c2..e43e124f4a 100644 --- a/test/registered/attention/test_qwen35_deterministic.py +++ b/test/registered/attention/test_qwen35_deterministic.py @@ -12,7 +12,7 @@ from sglang.test.test_deterministic_utils import ( TestDeterministicBase, ) -register_cuda_ci(est_time=119, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=187, stage="extra-b", runner_config="4-gpu-h100") QWEN35 = "Qwen/Qwen3.5-35B-A3B" diff --git a/test/registered/attention/test_triton_attention_kernels.py b/test/registered/attention/test_triton_attention_kernels.py index 347e77b14d..b09e63c4cf 100644 --- a/test/registered/attention/test_triton_attention_kernels.py +++ b/test/registered/attention/test_triton_attention_kernels.py @@ -25,7 +25,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase, is_in_amd_ci # Triton attention kernel unit tests (decode, extend, prefill) -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=36, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=30, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/attention/test_trtllm_mha_encoder_only.py b/test/registered/attention/test_trtllm_mha_encoder_only.py index 831b99de49..1aa988202c 100644 --- a/test/registered/attention/test_trtllm_mha_encoder_only.py +++ b/test/registered/attention/test_trtllm_mha_encoder_only.py @@ -17,7 +17,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase # trtllm_mha kernels are sm100-only; run this kernel-unit test on Blackwell. -register_cuda_ci(est_time=10, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=23, stage="base-b", runner_config="4-gpu-b200") DEVICE = "cuda" PAGE_SIZE = 32 diff --git a/test/registered/attention/test_trtllm_mha_page_table.py b/test/registered/attention/test_trtllm_mha_page_table.py index c70e852229..31e7f7563e 100644 --- a/test/registered/attention/test_trtllm_mha_page_table.py +++ b/test/registered/attention/test_trtllm_mha_page_table.py @@ -25,7 +25,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase # Triton kernel unit test for the trtllm_mha device-side page-table build. -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=14, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/attention/test_verify_splitkv.py b/test/registered/attention/test_verify_splitkv.py index 856d8e8a99..46bb162c3a 100644 --- a/test/registered/attention/test_verify_splitkv.py +++ b/test/registered/attention/test_verify_splitkv.py @@ -30,7 +30,7 @@ from sglang.kernels.ops.attention.verify_splitkv import ( from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=30, suite="stage-b-test-1-gpu-small-amd-mi35x") # Split-KV accumulates the prefix in parallel splits and merges via log-sum-exp; diff --git a/test/registered/attention/unittests/dense/test_extend_init_contract.py b/test/registered/attention/unittests/dense/test_extend_init_contract.py index 177012a732..6bcfa66db5 100644 --- a/test/registered/attention/unittests/dense/test_extend_init_contract.py +++ b/test/registered/attention/unittests/dense/test_extend_init_contract.py @@ -36,7 +36,7 @@ from sglang.test.kits.attention_unittest.attention_methods.dense_attention impor ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="base-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=14, stage="base-a", runner_config="1-gpu-small") _EXTEND_CASE = DenseAttentionCase( name="extend_no_prefix_smoke", diff --git a/test/registered/attention/unittests/dense/test_fa3.py b/test/registered/attention/unittests/dense/test_fa3.py index 2c77f986f4..38f6b91fb3 100644 --- a/test/registered/attention/unittests/dense/test_fa3.py +++ b/test/registered/attention/unittests/dense/test_fa3.py @@ -30,7 +30,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=32, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/dense/test_fa4.py b/test/registered/attention/unittests/dense/test_fa4.py index 2adbce5d6e..55017f52b7 100644 --- a/test/registered/attention/unittests/dense/test_fa4.py +++ b/test/registered/attention/unittests/dense/test_fa4.py @@ -29,8 +29,8 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=27, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=22, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=37, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=35, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/dense/test_flashinfer.py b/test/registered/attention/unittests/dense/test_flashinfer.py index 214bcecb74..f5b3d5cd71 100644 --- a/test/registered/attention/unittests/dense/test_flashinfer.py +++ b/test/registered/attention/unittests/dense/test_flashinfer.py @@ -26,8 +26,8 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=18, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=33, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=36, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf( diff --git a/test/registered/attention/unittests/dense/test_flex_attention.py b/test/registered/attention/unittests/dense/test_flex_attention.py index 1232433685..2f82af7be0 100644 --- a/test/registered/attention/unittests/dense/test_flex_attention.py +++ b/test/registered/attention/unittests/dense/test_flex_attention.py @@ -14,7 +14,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=18, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=30, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-large") diff --git a/test/registered/attention/unittests/dense/test_hybrid_attn.py b/test/registered/attention/unittests/dense/test_hybrid_attn.py index 444dd5a412..b8b2ff5257 100644 --- a/test/registered/attention/unittests/dense/test_hybrid_attn.py +++ b/test/registered/attention/unittests/dense/test_hybrid_attn.py @@ -18,8 +18,8 @@ from sglang.test.kits.attention_unittest.attention_methods.dense_attention impor ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=25, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=21, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf( diff --git a/test/registered/attention/unittests/dense/test_tbo.py b/test/registered/attention/unittests/dense/test_tbo.py index 90a49fae66..85abe233fe 100644 --- a/test/registered/attention/unittests/dense/test_tbo.py +++ b/test/registered/attention/unittests/dense/test_tbo.py @@ -21,8 +21,8 @@ from sglang.test.kits.attention_unittest.runner_modes.speculative_target_verify_ ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=22, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/dense/test_torch_native.py b/test/registered/attention/unittests/dense/test_torch_native.py index 9eebe7c4c6..7e23b8cc9f 100644 --- a/test/registered/attention/unittests/dense/test_torch_native.py +++ b/test/registered/attention/unittests/dense/test_torch_native.py @@ -11,7 +11,7 @@ from sglang.test.kits.attention_unittest.attention_methods.dense_attention impor ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=26, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/dense/test_triton.py b/test/registered/attention/unittests/dense/test_triton.py index 64559579ca..50d3d4d132 100644 --- a/test/registered/attention/unittests/dense/test_triton.py +++ b/test/registered/attention/unittests/dense/test_triton.py @@ -30,7 +30,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=18, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=30, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=19, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=25, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/dense/test_trtllm_mha.py b/test/registered/attention/unittests/dense/test_trtllm_mha.py index 26678d9b26..a0d2f37c72 100644 --- a/test/registered/attention/unittests/dense/test_trtllm_mha.py +++ b/test/registered/attention/unittests/dense/test_trtllm_mha.py @@ -28,8 +28,8 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=17, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=32, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=88, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf( diff --git a/test/registered/attention/unittests/dsa/test_dsa.py b/test/registered/attention/unittests/dsa/test_dsa.py index 3c63513ce8..e44a49251c 100644 --- a/test/registered/attention/unittests/dsa/test_dsa.py +++ b/test/registered/attention/unittests/dsa/test_dsa.py @@ -30,8 +30,8 @@ from sglang.test.kits.attention_unittest.runner_modes.speculative_draft_runner i ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=14, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=29, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=24, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/dsv4/test_deepseek_v4.py b/test/registered/attention/unittests/dsv4/test_deepseek_v4.py index 82d4142b9f..35572e52db 100644 --- a/test/registered/attention/unittests/dsv4/test_deepseek_v4.py +++ b/test/registered/attention/unittests/dsv4/test_deepseek_v4.py @@ -47,8 +47,8 @@ from sglang.test.kits.attention_unittest.runner_modes.speculative_target_verify_ run_dsv4_eagle_verify_cuda_graph_case, ) -register_cuda_ci(est_time=14, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=28, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=24, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/gdn/test_flashinfer.py b/test/registered/attention/unittests/gdn/test_flashinfer.py index 06de49b821..d87103ea56 100644 --- a/test/registered/attention/unittests/gdn/test_flashinfer.py +++ b/test/registered/attention/unittests/gdn/test_flashinfer.py @@ -25,8 +25,8 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=15, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=24, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=31, stage="base-b", runner_config="1-gpu-large") _cuda_major = int(torch.version.cuda.split(".")[0]) if torch.version.cuda else 0 _sm_major = torch.cuda.get_device_capability()[0] if torch.cuda.is_available() else 0 diff --git a/test/registered/attention/unittests/gdn/test_gdn_replayssm_spec_fold.py b/test/registered/attention/unittests/gdn/test_gdn_replayssm_spec_fold.py index f5e531e228..34c1256933 100644 --- a/test/registered/attention/unittests/gdn/test_gdn_replayssm_spec_fold.py +++ b/test/registered/attention/unittests/gdn/test_gdn_replayssm_spec_fold.py @@ -20,7 +20,7 @@ from sglang.kernels.ops.attention.fla.gdn_replayssm_spec_fold import ( from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=19, stage="base-b", runner_config="1-gpu-large") B, T = 3, 4 H, HV = 4, 8 diff --git a/test/registered/attention/unittests/gdn/test_torch_native.py b/test/registered/attention/unittests/gdn/test_torch_native.py index ef516ab5b7..cd5aac1d06 100644 --- a/test/registered/attention/unittests/gdn/test_torch_native.py +++ b/test/registered/attention/unittests/gdn/test_torch_native.py @@ -15,7 +15,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=16, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/gdn/test_triton.py b/test/registered/attention/unittests/gdn/test_triton.py index 129e76a5c8..39ccebfad3 100644 --- a/test/registered/attention/unittests/gdn/test_triton.py +++ b/test/registered/attention/unittests/gdn/test_triton.py @@ -29,7 +29,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=13, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=18, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/kda/test_triton.py b/test/registered/attention/unittests/kda/test_triton.py index d9f24bbabd..5b984221b5 100644 --- a/test/registered/attention/unittests/kda/test_triton.py +++ b/test/registered/attention/unittests/kda/test_triton.py @@ -22,7 +22,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=15, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=23, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/lightning/test_triton.py b/test/registered/attention/unittests/lightning/test_triton.py index 961014bb13..c1967be12a 100644 --- a/test/registered/attention/unittests/lightning/test_triton.py +++ b/test/registered/attention/unittests/lightning/test_triton.py @@ -19,7 +19,7 @@ from sglang.test.kits.attention_unittest.runner_modes.speculative_target_verify_ ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=17, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/mamba/test_mamba2.py b/test/registered/attention/unittests/mamba/test_mamba2.py index e635874c2b..653c88be79 100644 --- a/test/registered/attention/unittests/mamba/test_mamba2.py +++ b/test/registered/attention/unittests/mamba/test_mamba2.py @@ -32,7 +32,7 @@ from sglang.test.kits.attention_unittest.runner_modes.speculative_target_verify_ ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=13, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=26, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large") diff --git a/test/registered/attention/unittests/mamba/test_replay_state_indices_validator.py b/test/registered/attention/unittests/mamba/test_replay_state_indices_validator.py index b2f734e259..5f51759887 100644 --- a/test/registered/attention/unittests/mamba/test_replay_state_indices_validator.py +++ b/test/registered/attention/unittests/mamba/test_replay_state_indices_validator.py @@ -7,7 +7,7 @@ from sglang.srt.layers.attention.mamba.replay_state_indices_validator import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") class TestReplayStateIndicesValidator(unittest.TestCase): diff --git a/test/registered/attention/unittests/mla/test_flashinfer.py b/test/registered/attention/unittests/mla/test_flashinfer.py index 7005ebbf49..b76b016101 100644 --- a/test/registered/attention/unittests/mla/test_flashinfer.py +++ b/test/registered/attention/unittests/mla/test_flashinfer.py @@ -32,8 +32,8 @@ MLA_SHAPE_KWARGS = dict( from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=28, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=25, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/mla/test_flashmla.py b/test/registered/attention/unittests/mla/test_flashmla.py index 2e63fc781c..b4dfddf9a1 100644 --- a/test/registered/attention/unittests/mla/test_flashmla.py +++ b/test/registered/attention/unittests/mla/test_flashmla.py @@ -58,8 +58,8 @@ _DECODE_SKIP_REASON = ( from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=27, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=26, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/mla/test_triton.py b/test/registered/attention/unittests/mla/test_triton.py index 5c8956506c..795c39d74a 100644 --- a/test/registered/attention/unittests/mla/test_triton.py +++ b/test/registered/attention/unittests/mla/test_triton.py @@ -28,8 +28,8 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=18, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=33, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=36, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf(not torch.cuda.is_available(), "CUDA is required") diff --git a/test/registered/attention/unittests/swa/test_flashinfer.py b/test/registered/attention/unittests/swa/test_flashinfer.py index 23496cf5f2..c365cb5ff5 100644 --- a/test/registered/attention/unittests/swa/test_flashinfer.py +++ b/test/registered/attention/unittests/swa/test_flashinfer.py @@ -24,8 +24,8 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=13, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=28, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=24, stage="base-b", runner_config="1-gpu-large") @unittest.skipIf( diff --git a/test/registered/attention/unittests/swa/test_swa_out_cache_loc.py b/test/registered/attention/unittests/swa/test_swa_out_cache_loc.py index 0d0317c5a5..ddea46f43e 100644 --- a/test/registered/attention/unittests/swa/test_swa_out_cache_loc.py +++ b/test/registered/attention/unittests/swa/test_swa_out_cache_loc.py @@ -24,7 +24,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestSWAKVPoolSetKVBuffer(CustomTestCase): diff --git a/test/registered/attention/unittests/swa/test_torch_native.py b/test/registered/attention/unittests/swa/test_torch_native.py index f98a621001..877e0215f6 100644 --- a/test/registered/attention/unittests/swa/test_torch_native.py +++ b/test/registered/attention/unittests/swa/test_torch_native.py @@ -12,7 +12,7 @@ from sglang.test.kits.attention_unittest.attention_methods.dense_attention impor ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=19, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/attention/unittests/swa/test_triton.py b/test/registered/attention/unittests/swa/test_triton.py index aeeb84c085..aa47ae114b 100644 --- a/test/registered/attention/unittests/swa/test_triton.py +++ b/test/registered/attention/unittests/swa/test_triton.py @@ -23,7 +23,7 @@ from sglang.test.kits.attention_unittest.runner_modes.split_op_runner import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=16, stage="base-b", runner_config="4-gpu-b200") register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-large") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/backends/test_flashinfer_fusion_preflight.py b/test/registered/backends/test_flashinfer_fusion_preflight.py index f99b931c37..3305644ee3 100644 --- a/test/registered/backends/test_flashinfer_fusion_preflight.py +++ b/test/registered/backends/test_flashinfer_fusion_preflight.py @@ -11,7 +11,7 @@ from sglang.srt.utils import get_cuda_driver_bindings, is_flashinfer_available from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=28, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=33, stage="base-b", runner_config="2-gpu-large") WORLD_SIZE = 2 diff --git a/test/registered/bench_fn/test_bench_serving_reasoning_stream.py b/test/registered/bench_fn/test_bench_serving_reasoning_stream.py index a4c4096d3c..46a8e47d78 100644 --- a/test/registered/bench_fn/test_bench_serving_reasoning_stream.py +++ b/test/registered/bench_fn/test_bench_serving_reasoning_stream.py @@ -26,7 +26,7 @@ from sglang.benchmark.serving import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=16, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _free_port() -> int: diff --git a/test/registered/bench_fn/test_benchmark_datasets_api.py b/test/registered/bench_fn/test_benchmark_datasets_api.py index 8cf03b5053..a9412c0b12 100644 --- a/test/registered/bench_fn/test_benchmark_datasets_api.py +++ b/test/registered/bench_fn/test_benchmark_datasets_api.py @@ -60,7 +60,7 @@ from sglang.benchmark.serving import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=38, suite="base-a-test-cpu") +register_cpu_ci(est_time=14, suite="base-a-test-cpu") register_cpu_ci(est_time=46, suite="stage-b-test-cpu-intel") diff --git a/test/registered/bench_fn/test_steady_state_benchmark.py b/test/registered/bench_fn/test_steady_state_benchmark.py index fe02fa1dde..534e342217 100644 --- a/test/registered/bench_fn/test_steady_state_benchmark.py +++ b/test/registered/bench_fn/test_steady_state_benchmark.py @@ -13,7 +13,7 @@ from sglang.benchmark.steady_state import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") class _StringTokenizer: diff --git a/test/registered/chunked_prefill/test_mm_chunked_embedding_unit.py b/test/registered/chunked_prefill/test_mm_chunked_embedding_unit.py index fa10635060..79c8ff4d4f 100644 --- a/test/registered/chunked_prefill/test_mm_chunked_embedding_unit.py +++ b/test/registered/chunked_prefill/test_mm_chunked_embedding_unit.py @@ -25,7 +25,7 @@ from sglang.srt.multimodal.transport.cuda_ipc import ( from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") @pytest.fixture(autouse=True) diff --git a/test/registered/chunked_prefill/test_scripted_core_1gpu.py b/test/registered/chunked_prefill/test_scripted_core_1gpu.py index e37010d756..7df425a554 100644 --- a/test/registered/chunked_prefill/test_scripted_core_1gpu.py +++ b/test/registered/chunked_prefill/test_scripted_core_1gpu.py @@ -10,7 +10,7 @@ from sglang.test.scripted_runtime_chunked_helpers import ( run_until_finished, ) -register_cuda_ci(est_time=106, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=112, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=198, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/chunked_prefill/test_scripted_core_4gpu.py b/test/registered/chunked_prefill/test_scripted_core_4gpu.py index ff3f021db6..d42ca8e5d9 100644 --- a/test/registered/chunked_prefill/test_scripted_core_4gpu.py +++ b/test/registered/chunked_prefill/test_scripted_core_4gpu.py @@ -11,7 +11,7 @@ from sglang.test.scripted_runtime_chunked_helpers import ( run_until_finished, ) -register_cuda_ci(est_time=55, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=56, stage="extra-b", runner_config="4-gpu-h100") _CHUNK_SIZE = 64 diff --git a/test/registered/constrained_decoding/test_constrained_decoding.py b/test/registered/constrained_decoding/test_constrained_decoding.py index 3de939bcf4..08e622077c 100644 --- a/test/registered/constrained_decoding/test_constrained_decoding.py +++ b/test/registered/constrained_decoding/test_constrained_decoding.py @@ -17,7 +17,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=161, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=190, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=220, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/core/test_basic_sanity.py b/test/registered/core/test_basic_sanity.py index 4dca716259..57dac89714 100644 --- a/test/registered/core/test_basic_sanity.py +++ b/test/registered/core/test_basic_sanity.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=97, stage="base-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=121, stage="base-a", runner_config="1-gpu-small") register_amd_ci(est_time=160, suite="stage-a-test-1-gpu-small-amd") diff --git a/test/registered/core/test_basic_sanity_dflash.py b/test/registered/core/test_basic_sanity_dflash.py index d370856775..c696db2e2c 100644 --- a/test/registered/core/test_basic_sanity_dflash.py +++ b/test/registered/core/test_basic_sanity_dflash.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=135, stage="base-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=102, stage="base-a", runner_config="1-gpu-small") class TestBasicSanityDFlash( diff --git a/test/registered/core/test_basic_sanity_dspark.py b/test/registered/core/test_basic_sanity_dspark.py index bc0d55a27f..3675439ebd 100644 --- a/test/registered/core/test_basic_sanity_dspark.py +++ b/test/registered/core/test_basic_sanity_dspark.py @@ -16,7 +16,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=97, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=142, stage="base-b", runner_config="1-gpu-large") TARGET_MODEL = "Qwen/Qwen3-14B" DRAFT_MODEL = "deepseek-ai/dspark_qwen3_14b_block7" diff --git a/test/registered/core/test_basic_sanity_eagle3.py b/test/registered/core/test_basic_sanity_eagle3.py index 1ad5185a10..379201d213 100644 --- a/test/registered/core/test_basic_sanity_eagle3.py +++ b/test/registered/core/test_basic_sanity_eagle3.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=167, stage="base-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=131, stage="base-a", runner_config="1-gpu-small") register_amd_ci(est_time=200, suite="stage-a-test-1-gpu-small-amd") diff --git a/test/registered/core/test_engine_child_pids.py b/test/registered/core/test_engine_child_pids.py index 5b38585c58..8a2ff9b098 100644 --- a/test/registered/core/test_engine_child_pids.py +++ b/test/registered/core/test_engine_child_pids.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( CustomTestCase, ) -register_cuda_ci(est_time=77, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=49, stage="base-b", runner_config="1-gpu-small") class TestEngineChildPids(CustomTestCase): diff --git a/test/registered/core/test_hidden_states.py b/test/registered/core/test_hidden_states.py index 86d2a047fd..f2d362e67f 100644 --- a/test/registered/core/test_hidden_states.py +++ b/test/registered/core/test_hidden_states.py @@ -8,7 +8,7 @@ from sglang.srt.utils import get_device, is_hip from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST, CustomTestCase -register_cuda_ci(est_time=32, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=61, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=55, suite="stage-b-test-1-gpu-small-amd") _is_hip = is_hip() diff --git a/test/registered/core/test_no_extra_forked_cuda_context.py b/test/registered/core/test_no_extra_forked_cuda_context.py index 21eafda787..7049517984 100644 --- a/test/registered/core/test_no_extra_forked_cuda_context.py +++ b/test/registered/core/test_no_extra_forked_cuda_context.py @@ -15,7 +15,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=54, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=55, stage="base-b", runner_config="2-gpu-large") class TestTPServerGPUProcesses(CustomTestCase): diff --git a/test/registered/core/test_request_queue_validation.py b/test/registered/core/test_request_queue_validation.py index 6bd35a6366..1341568a1b 100644 --- a/test/registered/core/test_request_queue_validation.py +++ b/test/registered/core/test_request_queue_validation.py @@ -17,7 +17,7 @@ from sglang.test.test_utils import ( send_generate_requests, ) -register_cuda_ci(est_time=53, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=76, stage="base-b", runner_config="1-gpu-small") class TestMaxQueuedRequests(CustomTestCase): diff --git a/test/registered/core/test_srt_empty_deps.py b/test/registered/core/test_srt_empty_deps.py index 192a8b7d9c..967206690b 100644 --- a/test/registered/core/test_srt_empty_deps.py +++ b/test/registered/core/test_srt_empty_deps.py @@ -12,7 +12,7 @@ import pytest from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") # Packages known to transitively depend on torch or triton. # If a new package is added to runtime_base and it pulls torch, diff --git a/test/registered/core/test_srt_endpoint.py b/test/registered/core/test_srt_endpoint.py index d43465e756..b7b1ce921c 100644 --- a/test/registered/core/test_srt_endpoint.py +++ b/test/registered/core/test_srt_endpoint.py @@ -31,7 +31,7 @@ from sglang.test.test_utils import ( run_logprob_check, ) -register_cuda_ci(est_time=265, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=270, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=260, suite="stage-b-test-1-gpu-small-amd") SERVER_ENV = {"SGLANG_USE_PICKLE_IPC": "0"} diff --git a/test/registered/cp/test_cp_strategy_unit.py b/test/registered/cp/test_cp_strategy_unit.py index 70da8f1e58..925fe92910 100644 --- a/test/registered/cp/test_cp_strategy_unit.py +++ b/test/registered/cp/test_cp_strategy_unit.py @@ -41,7 +41,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _ExtendMode: diff --git a/test/registered/cp/test_deepseek_v4_flash_fp4_b200_cp.py b/test/registered/cp/test_deepseek_v4_flash_fp4_b200_cp.py index 8120531687..998c550eb8 100644 --- a/test/registered/cp/test_deepseek_v4_flash_fp4_b200_cp.py +++ b/test/registered/cp/test_deepseek_v4_flash_fp4_b200_cp.py @@ -22,7 +22,7 @@ from sglang.test.test_utils import ( try_cached_model, ) -register_cuda_ci(est_time=689, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=802, stage="extra-b", runner_config="4-gpu-b200") MODEL = "deepseek-ai/DeepSeek-V4-Flash" SERVER_LAUNCH_TIMEOUT = 3600 diff --git a/test/registered/cuda_graph/breakable/test_bcg_with_speculative_decoding.py b/test/registered/cuda_graph/breakable/test_bcg_with_speculative_decoding.py index b67bc159ad..760e1cac57 100644 --- a/test/registered/cuda_graph/breakable/test_bcg_with_speculative_decoding.py +++ b/test/registered/cuda_graph/breakable/test_bcg_with_speculative_decoding.py @@ -12,7 +12,7 @@ import unittest from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.server_fixtures.pcg_spec_fixture import PCGSpecBase -register_cuda_ci(est_time=106, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=110, stage="base-b", runner_config="2-gpu-large") class TestBCGWithEAGLE3(PCGSpecBase, unittest.TestCase): diff --git a/test/registered/dcp/test_dcp_layout_unit.py b/test/registered/dcp/test_dcp_layout_unit.py index 7c63381f2d..0c3da33f3c 100644 --- a/test/registered/dcp/test_dcp_layout_unit.py +++ b/test/registered/dcp/test_dcp_layout_unit.py @@ -33,7 +33,7 @@ from sglang.srt.mem_cache.memory_pool import HybridLinearKVPool from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") DCP_SIZES = [1, 2, 3, 4, 8] LENS = list(range(0, 41)) diff --git a/test/registered/dcp/test_trtllm_mla_family_dcp_metadata.py b/test/registered/dcp/test_trtllm_mla_family_dcp_metadata.py index 94de9c3db9..35df945733 100644 --- a/test/registered/dcp/test_trtllm_mla_family_dcp_metadata.py +++ b/test/registered/dcp/test_trtllm_mla_family_dcp_metadata.py @@ -23,7 +23,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=16, stage="base-b", runner_config="4-gpu-b200") NUM_DRAFT_TOKENS = 8 DCP_SIZE = 4 diff --git a/test/registered/debug_utils/test_tensor_dump_forward_hook.py b/test/registered/debug_utils/test_tensor_dump_forward_hook.py index 1d0d7b1147..b6e8ae5cb8 100644 --- a/test/registered/debug_utils/test_tensor_dump_forward_hook.py +++ b/test/registered/debug_utils/test_tensor_dump_forward_hook.py @@ -18,7 +18,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.layer_ut_utils import init_single_process_dist from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") # Backend-specific: the hook resolves get_device() and the matching distributed # backend, so only an AMD run exercises the HIP device and dump path. register_amd_ci(est_time=15, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/disaggregation/test_disaggregation_aarch64.py b/test/registered/disaggregation/test_disaggregation_aarch64.py index 9c6a2aa790..16305eabed 100644 --- a/test/registered/disaggregation/test_disaggregation_aarch64.py +++ b/test/registered/disaggregation/test_disaggregation_aarch64.py @@ -15,7 +15,7 @@ from sglang.test.test_utils import ( QWEN3_8B_MODEL_PATH = "Qwen/Qwen3-8B" -register_cuda_ci(est_time=125, stage="base-c", runner_config="4-gpu-gb300") +register_cuda_ci(est_time=223, stage="base-c", runner_config="4-gpu-gb300") # Keep rendezvous ports below the ephemeral range on the 4-GPU GB300 runner. NCCL_PORT_BASE = DEFAULT_PORT_FOR_SRT_TEST_RUNNER + 120 diff --git a/test/registered/disaggregation/test_disaggregation_basic.py b/test/registered/disaggregation/test_disaggregation_basic.py index 1764b7f037..299d0c218d 100644 --- a/test/registered/disaggregation/test_disaggregation_basic.py +++ b/test/registered/disaggregation/test_disaggregation_basic.py @@ -27,7 +27,7 @@ from sglang.test.test_utils import ( DEFAULT_TARGET_MODEL_EAGLE3, ) -register_cuda_ci(est_time=1007, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=671, stage="base-b", runner_config="2-gpu-large") class TestDisaggregationAccuracy(PauseResumeInPlaceMixin, PDDisaggregationServerBase): diff --git a/test/registered/disaggregation/test_disaggregation_chunked_prefill_abort.py b/test/registered/disaggregation/test_disaggregation_chunked_prefill_abort.py index 3550ca062d..fced3b2275 100644 --- a/test/registered/disaggregation/test_disaggregation_chunked_prefill_abort.py +++ b/test/registered/disaggregation/test_disaggregation_chunked_prefill_abort.py @@ -12,7 +12,7 @@ from sglang.test.server_fixtures.disaggregation_fixture import ( ) from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST -register_cuda_ci(est_time=120, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=82, stage="base-b", runner_config="2-gpu-large") PD_CHUNKED_ABORT_EXTRA_ARGS = [ "--max-running-requests", diff --git a/test/registered/disaggregation/test_disaggregation_decode_radix_cache_swa.py b/test/registered/disaggregation/test_disaggregation_decode_radix_cache_swa.py index 58ab206042..87e2d3645d 100644 --- a/test/registered/disaggregation/test_disaggregation_decode_radix_cache_swa.py +++ b/test/registered/disaggregation/test_disaggregation_decode_radix_cache_swa.py @@ -17,7 +17,7 @@ from sglang.test.server_fixtures.disaggregation_fixture import ( ) from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST_MXFP4_WITH_MOE, is_in_ci -register_cuda_ci(est_time=365, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=171, stage="extra-b", runner_config="8-gpu-h200") SWA_SERVER_ARGS = ["--page-size", "64", "--attention-backend", "triton"] diff --git a/test/registered/disaggregation/test_disaggregation_hisparse.py b/test/registered/disaggregation/test_disaggregation_hisparse.py index 1b212012b2..b5304281b3 100644 --- a/test/registered/disaggregation/test_disaggregation_hisparse.py +++ b/test/registered/disaggregation/test_disaggregation_hisparse.py @@ -12,7 +12,7 @@ from sglang.test.test_utils import ( try_cached_model, ) -register_cuda_ci(est_time=212, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=213, stage="extra-b", runner_config="8-gpu-h200") DSV4_FLASH_MODEL = "sgl-project/DeepSeek-V4-Flash-FP8" DSV4_FLASH_LOADER_CONFIG = '{"enable_multithread_load": true, "num_threads": 64}' diff --git a/test/registered/disaggregation/test_disaggregation_hybrid_attention.py b/test/registered/disaggregation/test_disaggregation_hybrid_attention.py index 66f19afee6..b188ba1aae 100644 --- a/test/registered/disaggregation/test_disaggregation_hybrid_attention.py +++ b/test/registered/disaggregation/test_disaggregation_hybrid_attention.py @@ -12,7 +12,7 @@ from sglang.test.test_utils import ( popen_launch_pd_server, ) -register_cuda_ci(est_time=578, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=575, stage="extra-b", runner_config="8-gpu-h200") @unittest.skipIf(is_in_ci(), "Temporarily disable the flaky test.") diff --git a/test/registered/disaggregation/test_disaggregation_inkling_mxfp8.py b/test/registered/disaggregation/test_disaggregation_inkling_mxfp8.py index 6821e91b8b..f30fa0b922 100644 --- a/test/registered/disaggregation/test_disaggregation_inkling_mxfp8.py +++ b/test/registered/disaggregation/test_disaggregation_inkling_mxfp8.py @@ -22,7 +22,7 @@ from sglang.test.test_utils import ( try_cached_model, ) -register_cuda_ci(est_time=447, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=624, stage="extra-b", runner_config="4-gpu-b200") MODEL = os.environ.get( "INKLING_SMALL_TEST_MODEL_PATH", "thinkingmachines/Inkling-Small-NVFP4" diff --git a/test/registered/disaggregation/test_disaggregation_rust_server.py b/test/registered/disaggregation/test_disaggregation_rust_server.py index 769a9fc79b..1d69f40790 100644 --- a/test/registered/disaggregation/test_disaggregation_rust_server.py +++ b/test/registered/disaggregation/test_disaggregation_rust_server.py @@ -27,7 +27,7 @@ from sglang.test.server_fixtures.disaggregation_fixture import ( ) from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST, is_rust_server_built -register_cuda_ci(est_time=66, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=92, stage="base-b", runner_config="2-gpu-large") @unittest.skipUnless( diff --git a/test/registered/dp_attn/test_dp_attention.py b/test/registered/dp_attn/test_dp_attention.py index 7e7ea2b6f6..0d707a7639 100644 --- a/test/registered/dp_attn/test_dp_attention.py +++ b/test/registered/dp_attn/test_dp_attention.py @@ -22,7 +22,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=443, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=330, stage="base-b", runner_config="2-gpu-large") register_amd_ci(est_time=500, suite="stage-b-test-2-gpu-large-amd") diff --git a/test/registered/dp_engine/test_data_parallelism.py b/test/registered/dp_engine/test_data_parallelism.py index c7b63fa3d1..f1fa5ddf97 100644 --- a/test/registered/dp_engine/test_data_parallelism.py +++ b/test/registered/dp_engine/test_data_parallelism.py @@ -14,7 +14,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=84, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=97, stage="base-b", runner_config="2-gpu-large") register_amd_ci(est_time=73, suite="stage-b-test-2-gpu-large-amd") diff --git a/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_swa.py b/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_swa.py index 7a5ced9751..4fb2b72a40 100644 --- a/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_swa.py +++ b/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_swa.py @@ -29,7 +29,7 @@ from sglang.test.server_fixtures.disaggregation_fixture import ( ) from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST_MXFP4_WITH_MOE -register_cuda_ci(est_time=1200, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=505, stage="extra-a", runner_config="2-gpu-large") UNIFIED_SWA_ARGS = [ "--skip-tokenizer-init", diff --git a/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_tri.py b/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_tri.py index 3e213c019b..96a37e4417 100644 --- a/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_tri.py +++ b/test/registered/e2e/disaggregation/test_disaggregation_unified_memory_tri.py @@ -32,7 +32,7 @@ from sglang.test.server_fixtures.disaggregation_fixture import ( PDDisaggregationServerBase, ) -register_cuda_ci(est_time=900, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=168, stage="extra-a", runner_config="2-gpu-large") _MODEL_PATH = os.environ.get("INKLING_TEST_MODEL_PATH", "thinkingmachines/Inkling") _MODEL_REVISION = os.environ.get("INKLING_TEST_MODEL_REVISION", "test") diff --git a/test/registered/e2e/hicache/test_hicache_unified_memory.py b/test/registered/e2e/hicache/test_hicache_unified_memory.py index ffccb24762..dc9459828d 100644 --- a/test/registered/e2e/hicache/test_hicache_unified_memory.py +++ b/test/registered/e2e/hicache/test_hicache_unified_memory.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=600, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=478, stage="extra-a", runner_config="2-gpu-large") _COMMON_ARGS = [ "--trust-remote-code", diff --git a/test/registered/e2e/layers/test_logprob_memory_distributed.py b/test/registered/e2e/layers/test_logprob_memory_distributed.py index cbeacc64ba..85cc5e0f47 100644 --- a/test/registered/e2e/layers/test_logprob_memory_distributed.py +++ b/test/registered/e2e/layers/test_logprob_memory_distributed.py @@ -15,7 +15,7 @@ from sglang.srt.layers.logprob_processor import InputLogprobProcessor from sglang.srt.model_executor.runner_utils import pool from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=60, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=8, stage="base-b", runner_config="2-gpu-large") def _run_rank(rank, rendezvous): diff --git a/test/registered/e2e/lora/test_lora_moe_prefill_cuda_graph.py b/test/registered/e2e/lora/test_lora_moe_prefill_cuda_graph.py index f43e3a5818..8cd0af9469 100644 --- a/test/registered/e2e/lora/test_lora_moe_prefill_cuda_graph.py +++ b/test/registered/e2e/lora/test_lora_moe_prefill_cuda_graph.py @@ -27,7 +27,7 @@ from sglang.test.lora_utils import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=600, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=159, stage="extra-a", runner_config="1-gpu-large") # Missing adapters shift logprobs by 7-17. LOGPROB_THRESHOLD = 1.0 diff --git a/test/registered/e2e/models/test_deepseek_v4_flash_fp4_megamoe_b200.py b/test/registered/e2e/models/test_deepseek_v4_flash_fp4_megamoe_b200.py index ac3f47d871..c9a8a9d7f3 100644 --- a/test/registered/e2e/models/test_deepseek_v4_flash_fp4_megamoe_b200.py +++ b/test/registered/e2e/models/test_deepseek_v4_flash_fp4_megamoe_b200.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( try_cached_model, ) -register_cuda_ci(est_time=283, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=340, stage="extra-b", runner_config="4-gpu-b200") MODEL = "deepseek-ai/DeepSeek-V4-Flash-0731" SERVER_LAUNCH_TIMEOUT = 3600 diff --git a/test/registered/e2e/models/test_generation_models.py b/test/registered/e2e/models/test_generation_models.py index 3812d3b296..9f4902fcdc 100644 --- a/test/registered/e2e/models/test_generation_models.py +++ b/test/registered/e2e/models/test_generation_models.py @@ -1,7 +1,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci # Generation model tests (CUDA only) -register_cuda_ci(est_time=144, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=229, stage="extra-a", runner_config="1-gpu-large") register_amd_ci(est_time=106, suite="stage-b-test-1-gpu-small-amd") # Copyright 2023-2024 SGLang Team diff --git a/test/registered/e2e/models/test_inkling_small_nvfp4.py b/test/registered/e2e/models/test_inkling_small_nvfp4.py index 41f7b8d4f4..49c6dac507 100644 --- a/test/registered/e2e/models/test_inkling_small_nvfp4.py +++ b/test/registered/e2e/models/test_inkling_small_nvfp4.py @@ -48,7 +48,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=1079, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=911, stage="extra-b", runner_config="4-gpu-b200") _MODEL_PATH = os.environ.get( "INKLING_SMALL_TEST_MODEL_PATH", "thinkingmachines/Inkling-Small-NVFP4" diff --git a/test/registered/e2e/models/test_kimi_k3_b300_low_latency.py b/test/registered/e2e/models/test_kimi_k3_b300_low_latency.py index e7e9e69cb9..5bc7526969 100644 --- a/test/registered/e2e/models/test_kimi_k3_b300_low_latency.py +++ b/test/registered/e2e/models/test_kimi_k3_b300_low_latency.py @@ -17,7 +17,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=1472, stage="extra-b", runner_config="8-gpu-b300") +register_cuda_ci(est_time=2287, stage="extra-b", runner_config="8-gpu-b300") MODEL_PATH = "moonshotai/Kimi-K3" DSPARK_DRAFT_MODEL = "RadixArk/Kimi-K3-DSpark" diff --git a/test/registered/e2e/models/test_kimi_linear_unified_memory_dcp_blackwell.py b/test/registered/e2e/models/test_kimi_linear_unified_memory_dcp_blackwell.py index e9f954cf8f..bb560a7f6d 100644 --- a/test/registered/e2e/models/test_kimi_linear_unified_memory_dcp_blackwell.py +++ b/test/registered/e2e/models/test_kimi_linear_unified_memory_dcp_blackwell.py @@ -47,7 +47,7 @@ from sglang.test.kits.eval_accuracy_kit import GSM8KMixin from sglang.test.kits.prefix_cache_branching_kit import PrefixCacheBranchingMixin from sglang.test.server_fixtures.default_fixture import DefaultServerBase -register_cuda_ci(est_time=161, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=233, stage="extra-b", runner_config="4-gpu-b200") KIMI_LINEAR_MODEL = "moonshotai/Kimi-Linear-48B-A3B-Instruct" diff --git a/test/registered/e2e/models/test_layernorm_sp.py b/test/registered/e2e/models/test_layernorm_sp.py index 29429ed506..5e6652aec0 100644 --- a/test/registered/e2e/models/test_layernorm_sp.py +++ b/test/registered/e2e/models/test_layernorm_sp.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=74, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=77, stage="base-b", runner_config="2-gpu-large") # Qwen3 dense (architecture "Qwen3ForCausalLM") is the SP allowlist entry. LAYERNORM_SP_MODEL = "Qwen/Qwen3-8B" diff --git a/test/registered/e2e/models/test_transformers_backend_eval.py b/test/registered/e2e/models/test_transformers_backend_eval.py index a2b1aa2ba8..20502d391e 100644 --- a/test/registered/e2e/models/test_transformers_backend_eval.py +++ b/test/registered/e2e/models/test_transformers_backend_eval.py @@ -11,7 +11,7 @@ from sglang.test.ci.ci_register import ( from sglang.test.few_shot_gsm8k import run_eval from sglang.test.server_fixtures.default_fixture import DefaultServerBase -register_cuda_ci(est_time=66, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=80, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=48, suite="stage-b-test-1-gpu-small-amd") register_cpu_ci(est_time=206, suite="stage-b-test-cpu-intel") diff --git a/test/registered/e2e/models/test_transformers_models.py b/test/registered/e2e/models/test_transformers_models.py index f8223f0545..0e76f3a792 100644 --- a/test/registered/e2e/models/test_transformers_models.py +++ b/test/registered/e2e/models/test_transformers_models.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=217, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=234, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=320, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/e2e/pp/test_pp_spec.py b/test/registered/e2e/pp/test_pp_spec.py index 264a6504b7..846368242d 100644 --- a/test/registered/e2e/pp/test_pp_spec.py +++ b/test/registered/e2e/pp/test_pp_spec.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=900, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=423, stage="extra-b", runner_config="4-gpu-h100") # topk=1 chains and a topk=2 tree: the relayed topology is constant for the # former and data-dependent for the latter, so both shapes are covered. diff --git a/test/registered/e2e/pp/test_pp_spec_embed_scan.py b/test/registered/e2e/pp/test_pp_spec_embed_scan.py index f42b5f85fa..96f4ef0fa9 100644 --- a/test/registered/e2e/pp/test_pp_spec_embed_scan.py +++ b/test/registered/e2e/pp/test_pp_spec_embed_scan.py @@ -16,7 +16,7 @@ from sglang.test.test_utils import ( publish_build_topology, ) -register_cuda_ci(est_time=60, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") # Shrunk from inclusionAI/Ling-mini-2.0; the field names are the real # checkpoint's, only the sizes are cut down for test speed. diff --git a/test/registered/e2e/speculative/test_dflash_domino.py b/test/registered/e2e/speculative/test_dflash_domino.py index 2c2a75f333..87c27c1102 100644 --- a/test/registered/e2e/speculative/test_dflash_domino.py +++ b/test/registered/e2e/speculative/test_dflash_domino.py @@ -14,7 +14,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=600, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=149, stage="base-b", runner_config="2-gpu-large") class TestDFlashDominoFullVocab(CustomTestCase): diff --git a/test/registered/ep/test_deepep_small_extra.py b/test/registered/ep/test_deepep_small_extra.py index 3b0a2c6b4a..6d40d3aec7 100644 --- a/test/registered/ep/test_deepep_small_extra.py +++ b/test/registered/ep/test_deepep_small_extra.py @@ -13,7 +13,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=283, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=287, stage="extra-b", runner_config="4-gpu-h100") class TestHybridDPTP(CustomTestCase): diff --git a/test/registered/ep/test_tbo_shared_experts_fusion.py b/test/registered/ep/test_tbo_shared_experts_fusion.py index b31ed171c6..c758f20f7c 100644 --- a/test/registered/ep/test_tbo_shared_experts_fusion.py +++ b/test/registered/ep/test_tbo_shared_experts_fusion.py @@ -12,7 +12,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=253, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=387, stage="extra-b", runner_config="8-gpu-h200") DEEPSEEK_V3_MODEL_PATH = "deepseek-ai/DeepSeek-V3-0324" diff --git a/test/registered/eplb/test_lplb_distributed.py b/test/registered/eplb/test_lplb_distributed.py index 5d9220d619..b5ac3d2bbc 100644 --- a/test/registered/eplb/test_lplb_distributed.py +++ b/test/registered/eplb/test_lplb_distributed.py @@ -40,7 +40,7 @@ from sglang.srt.distributed.parallel_state import ( from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import publish_build_topology -register_cuda_ci(est_time=18, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=24, stage="base-b", runner_config="2-gpu-large") NUM_GPUS = 2 TOPK = 2 diff --git a/test/registered/expert_pack/test_expert_pack_runtime.py b/test/registered/expert_pack/test_expert_pack_runtime.py index e611f70c56..a7ab1c69f5 100644 --- a/test/registered/expert_pack/test_expert_pack_runtime.py +++ b/test/registered/expert_pack/test_expert_pack_runtime.py @@ -13,7 +13,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") from sglang.srt.layers.moe.expert_pack import ( # noqa: E402 ExpertPackStore, diff --git a/test/registered/expert_pack/test_kimi_k3_gguf.py b/test/registered/expert_pack/test_kimi_k3_gguf.py index def8f2e097..85f89ded4a 100644 --- a/test/registered/expert_pack/test_kimi_k3_gguf.py +++ b/test/registered/expert_pack/test_kimi_k3_gguf.py @@ -20,7 +20,7 @@ from sglang.srt.model_loader.kimi_k3_gguf import ( from sglang.srt.models.kimi_k3 import _uses_split_gguf_kv_b from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestKimiK3GGUFMapping(unittest.TestCase): diff --git a/test/registered/function_call/test_kimik3_detector.py b/test/registered/function_call/test_kimik3_detector.py index 478dcc3ca4..47f2658414 100644 --- a/test/registered/function_call/test_kimik3_detector.py +++ b/test/registered/function_call/test_kimik3_detector.py @@ -16,7 +16,7 @@ from sglang.srt.function_call.kimik3_format import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_tool(name: str) -> Tool: diff --git a/test/registered/hicache/test_hicache_spec_file_storage.py b/test/registered/hicache/test_hicache_spec_file_storage.py index 1023531a7e..c9b73d6a5f 100644 --- a/test/registered/hicache/test_hicache_spec_file_storage.py +++ b/test/registered/hicache/test_hicache_spec_file_storage.py @@ -17,7 +17,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.hicache_spec_storage_common import HiCacheSpecStorageMixin from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=142, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=224, stage="extra-a", runner_config="1-gpu-large") @unittest.skipIf(is_hip(), "HiCache + EAGLE3 file-storage loadback e2e is CUDA-only.") diff --git a/test/registered/hicache/test_hicache_spec_mooncake_storage.py b/test/registered/hicache/test_hicache_spec_mooncake_storage.py index 804ded688a..31c0a7dc61 100644 --- a/test/registered/hicache/test_hicache_spec_mooncake_storage.py +++ b/test/registered/hicache/test_hicache_spec_mooncake_storage.py @@ -29,7 +29,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.hicache_spec_storage_common import HiCacheSpecStorageMixin from sglang.test.test_utils import CustomTestCase, find_available_port -register_cuda_ci(est_time=186, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=190, stage="extra-a", runner_config="2-gpu-large") @unittest.skipIf( diff --git a/test/registered/hicache/test_qwen35_hicache.py b/test/registered/hicache/test_qwen35_hicache.py index e42de0a6fe..f8a31a05ca 100644 --- a/test/registered/hicache/test_qwen35_hicache.py +++ b/test/registered/hicache/test_qwen35_hicache.py @@ -22,7 +22,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=369, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=295, stage="extra-b", runner_config="4-gpu-h100") QWEN35_27B_MODEL = "Qwen/Qwen3.5-27B" ACC_THRESHOLDS = {QWEN35_27B_MODEL: {"gsm8k": 0.8}} diff --git a/test/registered/kernels/benchmark/gemm/bench_kda_fp8_skinny_gemm.py b/test/registered/kernels/benchmark/gemm/bench_kda_fp8_skinny_gemm.py index cd8af5a868..fdab6ac62d 100644 --- a/test/registered/kernels/benchmark/gemm/bench_kda_fp8_skinny_gemm.py +++ b/test/registered/kernels/benchmark/gemm/bench_kda_fp8_skinny_gemm.py @@ -18,7 +18,7 @@ from sglang.srt.layers.quantization.fp8_utils import ( ) from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=90, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") def _make_inputs(m: int, n: int, k: int): diff --git a/test/registered/kernels/ops/attention/test_dcp_lse_combine.py b/test/registered/kernels/ops/attention/test_dcp_lse_combine.py index f177178c1f..1f9c75d8d9 100644 --- a/test/registered/kernels/ops/attention/test_dcp_lse_combine.py +++ b/test/registered/kernels/ops/attention/test_dcp_lse_combine.py @@ -16,7 +16,7 @@ import torch from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=23, stage="base-b", runner_config="1-gpu-large") class TestLSECombineTritonVsCPU(CustomTestCase): diff --git a/test/registered/kernels/ops/attention/test_dsa_indexer.py b/test/registered/kernels/ops/attention/test_dsa_indexer.py index f028468696..1cc537ff5f 100644 --- a/test/registered/kernels/ops/attention/test_dsa_indexer.py +++ b/test/registered/kernels/ops/attention/test_dsa_indexer.py @@ -34,7 +34,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch, ForwardMo from sglang.srt.server_args import ServerArgs, set_global_server_args_for_scheduler from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=31, stage="base-b", runner_config="1-gpu-large") # Global configuration for all indexer tests DEFAULT_CONFIG = { diff --git a/test/registered/kernels/ops/attention/test_sm120_paged_mqa_logits.py b/test/registered/kernels/ops/attention/test_sm120_paged_mqa_logits.py index 8b609b5612..23f71e8cba 100644 --- a/test/registered/kernels/ops/attention/test_sm120_paged_mqa_logits.py +++ b/test/registered/kernels/ops/attention/test_sm120_paged_mqa_logits.py @@ -29,7 +29,7 @@ from sglang.srt.layers.attention.dsv4.indexer import ( from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-small") # DSv4 indexer cache layout (fixed by deepseek_v4_memory_pool.DeepSeekV4IndexerPool): diff --git a/test/registered/kernels/ops/communication/test_symm_mem_all_gather.py b/test/registered/kernels/ops/communication/test_symm_mem_all_gather.py index 75f0aaf9e8..e999c2b128 100644 --- a/test/registered/kernels/ops/communication/test_symm_mem_all_gather.py +++ b/test/registered/kernels/ops/communication/test_symm_mem_all_gather.py @@ -35,7 +35,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.kernels.utils import multigpu_pytest_main -register_cuda_ci(est_time=38, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=32, stage="extra-b", runner_config="8-gpu-h200") # Nightly is not redundant here: it sets SGLANG_JIT_KERNEL_RUN_FULL_TESTS=1 to expand get_ci_test_range sweeps. register_cuda_ci(est_time=70, stage="nightly", runner_config="8-gpu-h200") diff --git a/test/registered/kernels/ops/diffusion/test_sites.py b/test/registered/kernels/ops/diffusion/test_sites.py index a625955294..109fff9ce6 100644 --- a/test/registered/kernels/ops/diffusion/test_sites.py +++ b/test/registered/kernels/ops/diffusion/test_sites.py @@ -49,7 +49,7 @@ from sglang.kernels.ops.diffusion import ( from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") register_cuda_ci(est_time=38, stage="base-b-kernel-unit", runner_config="1-gpu-large") requires_cuda = pytest.mark.skipif( diff --git a/test/registered/kernels/ops/embeddings/test_vocab_parallel_embedding.py b/test/registered/kernels/ops/embeddings/test_vocab_parallel_embedding.py index 45f9841d7e..6b9deeb523 100644 --- a/test/registered/kernels/ops/embeddings/test_vocab_parallel_embedding.py +++ b/test/registered/kernels/ops/embeddings/test_vocab_parallel_embedding.py @@ -14,7 +14,7 @@ from sglang.srt.layers.vocab_parallel_embedding import ( ) from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") pytestmark = pytest.mark.skipif( not torch.cuda.is_available(), reason="CUDA is required for this test." diff --git a/test/registered/kernels/ops/gemm/test_sm120_fp8_linear.py b/test/registered/kernels/ops/gemm/test_sm120_fp8_linear.py index 96cd27dbf0..c727b79ef1 100644 --- a/test/registered/kernels/ops/gemm/test_sm120_fp8_linear.py +++ b/test/registered/kernels/ops/gemm/test_sm120_fp8_linear.py @@ -11,7 +11,7 @@ from sglang.kernels.ops.gemm import try_sm120_fp8_linear from sglang.srt.layers.quantization.fp8_utils import apply_fp8_linear_bmm_flashinfer from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=180, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=34, stage="base-b", runner_config="1-gpu-small") if not (torch.cuda.is_available() and torch.cuda.get_device_capability() == (12, 0)): pytest.skip( diff --git a/test/registered/kernels/ops/kv_canary/test_utils.py b/test/registered/kernels/ops/kv_canary/test_utils.py index ea41ffea9d..206acd0015 100644 --- a/test/registered/kernels/ops/kv_canary/test_utils.py +++ b/test/registered/kernels/ops/kv_canary/test_utils.py @@ -8,7 +8,7 @@ from sglang.kernels.jit.benchmark.kv_canary.utils import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, stage="base-a", runner_config="cpu") +register_cpu_ci(est_time=1, stage="base-a", runner_config="cpu") def test_fast_matrix_cases_include_e2e_decode_and_chunked_prefill_scenarios() -> None: diff --git a/test/registered/kernels/ops/layernorm/test_mhc_kernels.py b/test/registered/kernels/ops/layernorm/test_mhc_kernels.py index b2010cd6ce..88c09bc902 100644 --- a/test/registered/kernels/ops/layernorm/test_mhc_kernels.py +++ b/test/registered/kernels/ops/layernorm/test_mhc_kernels.py @@ -8,7 +8,7 @@ import sglang.kernels.ops.layernorm.mhc as mhc from sglang.kernels.ops.layernorm.mhc import mhc_fused_post_pre, mhc_post, mhc_pre from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=29, stage="base-b", runner_config="1-gpu-large") @pytest.fixture diff --git a/test/registered/kernels/ops/speculative/test_boundary_kv_fix_kernels.py b/test/registered/kernels/ops/speculative/test_boundary_kv_fix_kernels.py index 6d48f4a111..97d938f5df 100644 --- a/test/registered/kernels/ops/speculative/test_boundary_kv_fix_kernels.py +++ b/test/registered/kernels/ops/speculative/test_boundary_kv_fix_kernels.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") """Boundary-KV fix kernels (SGLANG_ENABLE_MTP_BOUNDARY_KV_FIX) vs a pure-torch reference. diff --git a/test/registered/kv_canary/test_self_e2e_baseline.py b/test/registered/kv_canary/test_self_e2e_baseline.py index 500687723c..ca703e1ec3 100644 --- a/test/registered/kv_canary/test_self_e2e_baseline.py +++ b/test/registered/kv_canary/test_self_e2e_baseline.py @@ -7,7 +7,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.consts import SWA_POOL_SERVER_ARGS from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=135, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=161, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=236, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_e2e_bench_speed.py b/test/registered/kv_canary/test_self_e2e_bench_speed.py index e0e4d7e4d5..40560b47db 100644 --- a/test/registered/kv_canary/test_self_e2e_bench_speed.py +++ b/test/registered/kv_canary/test_self_e2e_bench_speed.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import DEFAULT_PORT_FOR_SRT_TEST_RUNNER # CUDA-only: this self-bench asserts a 1.0% kv_canary overhead budget tuned on # the CUDA (H100) runner. On ROCm the measured overhead is ~1.26%, so the # benchmark is not portable as-is; keep it off AMD CI rather than register-and-skip. -register_cuda_ci(est_time=304, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=743, stage="extra-a", runner_config="1-gpu-large") _QWEN3_MODEL = "Qwen/Qwen3-30B-A3B" diff --git a/test/registered/kv_canary/test_self_e2e_pd_baseline.py b/test/registered/kv_canary/test_self_e2e_pd_baseline.py index 902b912590..b3cd359806 100644 --- a/test/registered/kv_canary/test_self_e2e_pd_baseline.py +++ b/test/registered/kv_canary/test_self_e2e_pd_baseline.py @@ -5,7 +5,7 @@ import unittest from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.pd_fixture import CanaryPDFixture -register_cuda_ci(est_time=204, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=176, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=106, stage="extra-a", runner_config="2-gpu-large-amd") diff --git a/test/registered/kv_canary/test_self_e2e_pd_perturb.py b/test/registered/kv_canary/test_self_e2e_pd_perturb.py index c2ea1ff0a0..60ba7dcfe7 100644 --- a/test/registered/kv_canary/test_self_e2e_pd_perturb.py +++ b/test/registered/kv_canary/test_self_e2e_pd_perturb.py @@ -7,7 +7,7 @@ from sglang.srt.kv_canary.perturb.config import TargetGroupKind from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.pd_fixture import CanaryPDFixture -register_cuda_ci(est_time=305, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=266, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=231, stage="extra-a", runner_config="2-gpu-large-amd") diff --git a/test/registered/kv_canary/test_self_e2e_perturb_raise.py b/test/registered/kv_canary/test_self_e2e_perturb_raise.py index fc66f0d2fe..58ccc6dfe6 100644 --- a/test/registered/kv_canary/test_self_e2e_perturb_raise.py +++ b/test/registered/kv_canary/test_self_e2e_perturb_raise.py @@ -7,7 +7,7 @@ from sglang.srt.kv_canary.perturb.config import TargetGroupKind from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=54, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=127, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=50, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_e2e_perturb_real_kv_unused_cache.py b/test/registered/kv_canary/test_self_e2e_perturb_real_kv_unused_cache.py index dd9cbebea1..79e920fdd0 100644 --- a/test/registered/kv_canary/test_self_e2e_perturb_real_kv_unused_cache.py +++ b/test/registered/kv_canary/test_self_e2e_perturb_real_kv_unused_cache.py @@ -9,7 +9,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.consts import SWA_POOL_SERVER_ARGS from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=446, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=345, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=503, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_e2e_perturb_real_kv_used.py b/test/registered/kv_canary/test_self_e2e_perturb_real_kv_used.py index d2c494e51c..f82d76ebd5 100644 --- a/test/registered/kv_canary/test_self_e2e_perturb_real_kv_used.py +++ b/test/registered/kv_canary/test_self_e2e_perturb_real_kv_used.py @@ -9,7 +9,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.consts import SWA_POOL_SERVER_ARGS from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=290, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=318, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=256, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_e2e_perturb_req_to_token.py b/test/registered/kv_canary/test_self_e2e_perturb_req_to_token.py index ce390bfd56..fff9086a27 100644 --- a/test/registered/kv_canary/test_self_e2e_perturb_req_to_token.py +++ b/test/registered/kv_canary/test_self_e2e_perturb_req_to_token.py @@ -7,7 +7,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.consts import SWA_POOL_SERVER_ARGS from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=194, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=218, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=175, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_e2e_pp_baseline.py b/test/registered/kv_canary/test_self_e2e_pp_baseline.py index 8b7fd1e976..daac86b8c8 100644 --- a/test/registered/kv_canary/test_self_e2e_pp_baseline.py +++ b/test/registered/kv_canary/test_self_e2e_pp_baseline.py @@ -6,7 +6,7 @@ from sglang.srt.kv_canary.config import CanaryMode from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.pp_fixture import CanaryPPFixture -register_cuda_ci(est_time=207, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=248, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=243, stage="extra-a", runner_config="2-gpu-large-amd") diff --git a/test/registered/kv_canary/test_self_e2e_pp_perturb.py b/test/registered/kv_canary/test_self_e2e_pp_perturb.py index 1a16bb4d05..846f9ad105 100644 --- a/test/registered/kv_canary/test_self_e2e_pp_perturb.py +++ b/test/registered/kv_canary/test_self_e2e_pp_perturb.py @@ -8,7 +8,7 @@ from sglang.srt.kv_canary.perturb.config import TargetGroupKind from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.pp_fixture import CanaryPPFixture -register_cuda_ci(est_time=318, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=365, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=298, stage="extra-a", runner_config="2-gpu-large-amd") diff --git a/test/registered/kv_canary/test_self_e2e_pr_25015.py b/test/registered/kv_canary/test_self_e2e_pr_25015.py index ea004b2f0c..e69cc33293 100644 --- a/test/registered/kv_canary/test_self_e2e_pr_25015.py +++ b/test/registered/kv_canary/test_self_e2e_pr_25015.py @@ -9,7 +9,7 @@ from sglang.srt.kv_canary.config import CanaryMode from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=102, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=128, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=101, stage="extra-a", runner_config="1-gpu-small-amd") _SPEC_EAGLE_TOKEN_ORACLE_ENV = { diff --git a/test/registered/kv_canary/test_self_e2e_pr_26329.py b/test/registered/kv_canary/test_self_e2e_pr_26329.py index f1b4e4d7b5..732b1f0286 100644 --- a/test/registered/kv_canary/test_self_e2e_pr_26329.py +++ b/test/registered/kv_canary/test_self_e2e_pr_26329.py @@ -11,7 +11,7 @@ from sglang.srt.kv_canary.config import CanaryMode from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.e2e_base import CanaryE2EBase -register_cuda_ci(est_time=105, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=124, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=99, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_buffer_alloc.py b/test/registered/kv_canary/test_self_unit_buffer_alloc.py index 91d21c8b00..dc3beb119e 100644 --- a/test/registered/kv_canary/test_self_unit_buffer_alloc.py +++ b/test/registered/kv_canary/test_self_unit_buffer_alloc.py @@ -15,7 +15,7 @@ from sglang.srt.kv_canary.pool_patcher.buffer_alloc import ( from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=10, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_capacities.py b/test/registered/kv_canary/test_self_unit_capacities.py index 31325b437b..c00d8f51c6 100644 --- a/test/registered/kv_canary/test_self_unit_capacities.py +++ b/test/registered/kv_canary/test_self_unit_capacities.py @@ -18,7 +18,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestComputeLaunchCapacities(CustomTestCase): diff --git a/test/registered/kv_canary/test_self_unit_capture_refusal.py b/test/registered/kv_canary/test_self_unit_capture_refusal.py index 84a1009c2a..85b84d446e 100644 --- a/test/registered/kv_canary/test_self_unit_capture_refusal.py +++ b/test/registered/kv_canary/test_self_unit_capture_refusal.py @@ -16,7 +16,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class TestTorchReferenceConflictsWithDecodeGraph(CustomTestCase): diff --git a/test/registered/kv_canary/test_self_unit_endpoint.py b/test/registered/kv_canary/test_self_unit_endpoint.py index 95163abaa6..6356514951 100644 --- a/test/registered/kv_canary/test_self_unit_endpoint.py +++ b/test/registered/kv_canary/test_self_unit_endpoint.py @@ -25,7 +25,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.fixtures import DEFAULT_DEVICE from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=20, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_future_tensor.py b/test/registered/kv_canary/test_self_unit_future_tensor.py index 6b6cb6e66e..ea7c3c5913 100644 --- a/test/registered/kv_canary/test_self_unit_future_tensor.py +++ b/test/registered/kv_canary/test_self_unit_future_tensor.py @@ -14,7 +14,7 @@ from sglang.test.ci.ci_register import ( from sglang.test.kv_canary.fixtures import DEFAULT_DEVICE from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=20, suite="extra-a-test-1-gpu-small-amd") register_xpu_ci(est_time=20, suite="stage-b-test-1-gpu-xpu") diff --git a/test/registered/kv_canary/test_self_unit_perturb.py b/test/registered/kv_canary/test_self_unit_perturb.py index 10a21dc547..1c87696384 100644 --- a/test/registered/kv_canary/test_self_unit_perturb.py +++ b/test/registered/kv_canary/test_self_unit_perturb.py @@ -40,7 +40,7 @@ from sglang.test.test_utils import CustomTestCase if TYPE_CHECKING: from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=14, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=10, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_plan_input.py b/test/registered/kv_canary/test_self_unit_plan_input.py index 98073e2ef8..03dac3acd3 100644 --- a/test/registered/kv_canary/test_self_unit_plan_input.py +++ b/test/registered/kv_canary/test_self_unit_plan_input.py @@ -18,7 +18,7 @@ from sglang.test.kv_canary.fixtures import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=30, suite="extra-a-test-1-gpu-small-amd") register_xpu_ci(est_time=30, suite="stage-b-test-1-gpu-xpu") diff --git a/test/registered/kv_canary/test_self_unit_pool_patcher.py b/test/registered/kv_canary/test_self_unit_pool_patcher.py index 5dc608479d..5d43f83e4c 100644 --- a/test/registered/kv_canary/test_self_unit_pool_patcher.py +++ b/test/registered/kv_canary/test_self_unit_pool_patcher.py @@ -24,7 +24,7 @@ from sglang.test.kv_canary.fixtures import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=45, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_pool_patcher_utils.py b/test/registered/kv_canary/test_self_unit_pool_patcher_utils.py index ea4012f59e..c48da2831c 100644 --- a/test/registered/kv_canary/test_self_unit_pool_patcher_utils.py +++ b/test/registered/kv_canary/test_self_unit_pool_patcher_utils.py @@ -6,7 +6,7 @@ from sglang.srt.kv_canary.pool_patcher.utils import wrap_method from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=10, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_radix_walker.py b/test/registered/kv_canary/test_self_unit_radix_walker.py index 3d94ca4aed..2cce9e827c 100644 --- a/test/registered/kv_canary/test_self_unit_radix_walker.py +++ b/test/registered/kv_canary/test_self_unit_radix_walker.py @@ -19,7 +19,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.fixtures import DEFAULT_DEVICE, make_radix_cache from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=30, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_req_to_expected_token_ids_manager.py b/test/registered/kv_canary/test_self_unit_req_to_expected_token_ids_manager.py index 575336628c..737a5deaa0 100644 --- a/test/registered/kv_canary/test_self_unit_req_to_expected_token_ids_manager.py +++ b/test/registered/kv_canary/test_self_unit_req_to_expected_token_ids_manager.py @@ -22,7 +22,7 @@ from sglang.test.kv_canary.fixtures import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=14, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=15, suite="extra-a-test-1-gpu-small-amd") register_xpu_ci(est_time=30, suite="stage-b-test-1-gpu-xpu") diff --git a/test/registered/kv_canary/test_self_unit_runner_health.py b/test/registered/kv_canary/test_self_unit_runner_health.py index 1ee396ea47..36c356f868 100644 --- a/test/registered/kv_canary/test_self_unit_runner_health.py +++ b/test/registered/kv_canary/test_self_unit_runner_health.py @@ -19,7 +19,7 @@ from sglang.test.kv_canary.runner_test_base import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=45, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_runner_per_forward.py b/test/registered/kv_canary/test_self_unit_runner_per_forward.py index 0fc87917ed..42b8298ecb 100644 --- a/test/registered/kv_canary/test_self_unit_runner_per_forward.py +++ b/test/registered/kv_canary/test_self_unit_runner_per_forward.py @@ -20,7 +20,7 @@ from sglang.test.kv_canary.runner_test_base import ( make_manager, ) -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=14, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=45, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_runner_swa_divergence.py b/test/registered/kv_canary/test_self_unit_runner_swa_divergence.py index 2c1a350db9..a6676ddbd7 100644 --- a/test/registered/kv_canary/test_self_unit_runner_swa_divergence.py +++ b/test/registered/kv_canary/test_self_unit_runner_swa_divergence.py @@ -25,7 +25,7 @@ from sglang.test.kv_canary.fixtures import DEFAULT_DEVICE, make_buffer_group from sglang.test.kv_canary.runner_test_base import CanaryManagerTestCase, make_manager from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=45, suite="extra-a-test-1-gpu-small-amd") register_xpu_ci(est_time=60, suite="stage-b-test-1-gpu-xpu") diff --git a/test/registered/kv_canary/test_self_unit_runner_sweep.py b/test/registered/kv_canary/test_self_unit_runner_sweep.py index ee59aee511..e59305b2cd 100644 --- a/test/registered/kv_canary/test_self_unit_runner_sweep.py +++ b/test/registered/kv_canary/test_self_unit_runner_sweep.py @@ -16,7 +16,7 @@ from sglang.test.kv_canary.runner_test_base import ( make_manager, ) -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=45, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_sweep_plan_builder.py b/test/registered/kv_canary/test_self_unit_sweep_plan_builder.py index c4d973c833..4e2088689a 100644 --- a/test/registered/kv_canary/test_self_unit_sweep_plan_builder.py +++ b/test/registered/kv_canary/test_self_unit_sweep_plan_builder.py @@ -13,7 +13,7 @@ from sglang.test.kv_canary.fixtures import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=14, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=30, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_token_oracle.py b/test/registered/kv_canary/test_self_unit_token_oracle.py index 563db9138f..4982809816 100644 --- a/test/registered/kv_canary/test_self_unit_token_oracle.py +++ b/test/registered/kv_canary/test_self_unit_token_oracle.py @@ -13,7 +13,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.kv_canary.fixtures import DEFAULT_DEVICE from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=1, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/kv_canary/test_self_unit_violation.py b/test/registered/kv_canary/test_self_unit_violation.py index 66ca4bd031..a72536c57c 100644 --- a/test/registered/kv_canary/test_self_unit_violation.py +++ b/test/registered/kv_canary/test_self_unit_violation.py @@ -18,7 +18,7 @@ from sglang.srt.kv_canary.runner.violation_reporter import ( from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=5, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/layers/mamba/test_causal_conv1d.py b/test/registered/layers/mamba/test_causal_conv1d.py index 638da5dd0d..f2c225e150 100644 --- a/test/registered/layers/mamba/test_causal_conv1d.py +++ b/test/registered/layers/mamba/test_causal_conv1d.py @@ -2,7 +2,7 @@ # SPDX-FileCopyrightText: Copyright contributors to the vLLM project from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=25, suite="stage-b-test-1-gpu-small-amd") # Adapted from https://github.com/vllm-project/vllm/blob/main/tests/kernels/mamba/test_causal_conv1d.py diff --git a/test/registered/layers/mamba/test_mamba2_mixer.py b/test/registered/layers/mamba/test_mamba2_mixer.py index 4631185959..ea4f444cb7 100644 --- a/test/registered/layers/mamba/test_mamba2_mixer.py +++ b/test/registered/layers/mamba/test_mamba2_mixer.py @@ -19,7 +19,7 @@ from sglang.srt.utils import get_device, get_device_count from sglang.test.ci.ci_register import register_cuda_ci, register_xpu_ci from sglang.test.test_utils import publish_build_topology -register_cuda_ci(est_time=30, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=37, stage="base-b", runner_config="2-gpu-large") register_xpu_ci(est_time=60, suite="nightly-xpu-2-gpu", nightly=True) NUM_GPUS = 2 diff --git a/test/registered/layers/mamba/test_mamba_slot_fused.py b/test/registered/layers/mamba/test_mamba_slot_fused.py index 484518be42..a820e33a95 100644 --- a/test/registered/layers/mamba/test_mamba_slot_fused.py +++ b/test/registered/layers/mamba/test_mamba_slot_fused.py @@ -24,7 +24,7 @@ from sglang.srt.utils.common import get_device_module from sglang.test.ci.ci_register import register_cuda_ci, register_xpu_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") register_xpu_ci(est_time=20, suite="stage-b-test-1-gpu-xpu") # Backends these kernels are verified against; extend as others gain Triton. diff --git a/test/registered/layers/mamba/test_mamba_ssm.py b/test/registered/layers/mamba/test_mamba_ssm.py index 250c9ae538..66f3898666 100644 --- a/test/registered/layers/mamba/test_mamba_ssm.py +++ b/test/registered/layers/mamba/test_mamba_ssm.py @@ -2,7 +2,7 @@ # SPDX-FileCopyrightText: Copyright contributors to the vLLM project from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-small-amd") # Adapted from https://github.com/vllm-project/vllm/blob/633f943e30a4444d890d26b81850f7217736f840/tests/kernels/mamba/test_mamba_ssm_ssd.py diff --git a/test/registered/layers/mamba/test_mamba_ssm_ssd.py b/test/registered/layers/mamba/test_mamba_ssm_ssd.py index 6933ea72e8..7f5dad856b 100644 --- a/test/registered/layers/mamba/test_mamba_ssm_ssd.py +++ b/test/registered/layers/mamba/test_mamba_ssm_ssd.py @@ -2,7 +2,7 @@ # SPDX-FileCopyrightText: Copyright contributors to the vLLM project from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=23, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=34, suite="stage-b-test-1-gpu-small-amd") # Adapted from https://github.com/vllm-project/vllm/blob/633f943e30a4444d890d26b81850f7217736f840/tests/kernels/mamba/test_mamba_ssm_ssd.py diff --git a/test/registered/mem_cache/test_int8_checkpoint_store.py b/test/registered/mem_cache/test_int8_checkpoint_store.py index 7fa42d651f..b668ec4f2c 100644 --- a/test/registered/mem_cache/test_int8_checkpoint_store.py +++ b/test/registered/mem_cache/test_int8_checkpoint_store.py @@ -22,7 +22,7 @@ from sglang.test.ci.ci_register import register_cpu_ci # case self-skips when no GPU is present), so they run on the cheapest CI tier. # The int8 GPU decode path is covered end-to-end by # test_int8_mamba_checkpoint_e2e (now in the extra stage). -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") H, V, K = 32, 128, 128 L = 4 diff --git a/test/registered/mem_cache/test_post_capture_kv_sizing.py b/test/registered/mem_cache/test_post_capture_kv_sizing.py index bbe3445095..43cc1042fc 100644 --- a/test/registered/mem_cache/test_post_capture_kv_sizing.py +++ b/test/registered/mem_cache/test_post_capture_kv_sizing.py @@ -28,7 +28,7 @@ from sglang.test.test_utils import ( ) # CI Registration -register_cuda_ci(est_time=75, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=117, stage="base-b", runner_config="1-gpu-large") STDOUT_FILENAME = "post_capture_kv_sizing_stdout.log" STDERR_FILENAME = "post_capture_kv_sizing_stderr.log" diff --git a/test/registered/mock_model/test_e2e_cp.py b/test/registered/mock_model/test_e2e_cp.py index 700d22b3f6..7838af71ca 100644 --- a/test/registered/mock_model/test_e2e_cp.py +++ b/test/registered/mock_model/test_e2e_cp.py @@ -6,7 +6,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.mock_model.utils import run_mock_model_bench_serving from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=50, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=55, stage="extra-a", runner_config="2-gpu-large") class TestE2EContextParallel(CustomTestCase): diff --git a/test/registered/mock_model/test_e2e_pd.py b/test/registered/mock_model/test_e2e_pd.py index bb1c84f8cc..19420aa0bf 100644 --- a/test/registered/mock_model/test_e2e_pd.py +++ b/test/registered/mock_model/test_e2e_pd.py @@ -19,7 +19,7 @@ from sglang.test.server_fixtures.disaggregation_fixture import ( PDDisaggregationServerBase, ) -register_cuda_ci(est_time=266, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=178, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=165, stage="extra-a", runner_config="2-gpu-large-amd") # DO NOT pass --disable-cuda-graph in canary e2e tests. The canary kernel diff --git a/test/registered/mock_model/test_e2e_pp.py b/test/registered/mock_model/test_e2e_pp.py index 7b3bed9528..6f827afb26 100644 --- a/test/registered/mock_model/test_e2e_pp.py +++ b/test/registered/mock_model/test_e2e_pp.py @@ -6,7 +6,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.mock_model.utils import run_mock_model_bench_serving from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=67, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=70, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=67, stage="extra-a", runner_config="2-gpu-large-amd") diff --git a/test/registered/mock_model/test_e2e_spec_eagle.py b/test/registered/mock_model/test_e2e_spec_eagle.py index f04361bbc8..f23c405232 100644 --- a/test/registered/mock_model/test_e2e_spec_eagle.py +++ b/test/registered/mock_model/test_e2e_spec_eagle.py @@ -6,7 +6,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.mock_model.utils import MOCK_MODEL_PATH, run_mock_model_bench_serving from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=96, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=97, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=77, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/mock_model/test_self_e2e_perturb_next_token_swap.py b/test/registered/mock_model/test_self_e2e_perturb_next_token_swap.py index 28ee1fa395..49e9b4e0ff 100644 --- a/test/registered/mock_model/test_self_e2e_perturb_next_token_swap.py +++ b/test/registered/mock_model/test_self_e2e_perturb_next_token_swap.py @@ -5,7 +5,7 @@ import unittest from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.mock_model.perturb_e2e_base import MockModelPerturbE2EBase -register_cuda_ci(est_time=55, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=68, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=131, stage="extra-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/mock_model/test_self_unit_canary_mock_wiring.py b/test/registered/mock_model/test_self_unit_canary_mock_wiring.py index 84c0808d52..6ae3d07ebe 100644 --- a/test/registered/mock_model/test_self_unit_canary_mock_wiring.py +++ b/test/registered/mock_model/test_self_unit_canary_mock_wiring.py @@ -16,7 +16,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.mock_model.utils import mock_model_server_args, mock_model_server_env from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=60, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/mock_model/test_self_unit_install.py b/test/registered/mock_model/test_self_unit_install.py index a0a5204fca..ef70413c42 100644 --- a/test/registered/mock_model/test_self_unit_install.py +++ b/test/registered/mock_model/test_self_unit_install.py @@ -12,7 +12,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=60, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/mock_model/test_self_unit_oracle.py b/test/registered/mock_model/test_self_unit_oracle.py index d90bb7536a..7ead5bf098 100644 --- a/test/registered/mock_model/test_self_unit_oracle.py +++ b/test/registered/mock_model/test_self_unit_oracle.py @@ -13,7 +13,7 @@ from sglang.srt.kv_canary.token_oracle.oracle import ( from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=60, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/mock_model/test_self_unit_oracle_torch_vs_ref.py b/test/registered/mock_model/test_self_unit_oracle_torch_vs_ref.py index 6a4e6010d7..77e5347363 100644 --- a/test/registered/mock_model/test_self_unit_oracle_torch_vs_ref.py +++ b/test/registered/mock_model/test_self_unit_oracle_torch_vs_ref.py @@ -10,7 +10,7 @@ from sglang.srt.kv_canary.token_oracle.oracle import HashOracle from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=30, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/mock_model/test_self_unit_sampler_hookpoint.py b/test/registered/mock_model/test_self_unit_sampler_hookpoint.py index 987666c294..5f25849a92 100644 --- a/test/registered/mock_model/test_self_unit_sampler_hookpoint.py +++ b/test/registered/mock_model/test_self_unit_sampler_hookpoint.py @@ -21,7 +21,7 @@ from sglang.srt.server_args import SAMPLING_BACKEND_CHOICES from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=60, suite="extra-a-test-1-gpu-small-amd") diff --git a/test/registered/model_loading/test_external_models.py b/test/registered/model_loading/test_external_models.py index ac02cb23cb..b90be5e3cd 100644 --- a/test/registered/model_loading/test_external_models.py +++ b/test/registered/model_loading/test_external_models.py @@ -9,7 +9,7 @@ from sglang.test.ci.ci_register import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=34, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=39, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=45, suite="stage-b-test-1-gpu-small-amd") register_cpu_ci(est_time=32, suite="stage-b-test-cpu-intel") diff --git a/test/registered/model_loading/test_load_weights_from_remote_instance.py b/test/registered/model_loading/test_load_weights_from_remote_instance.py index a4e5a8966b..8ded73fb71 100644 --- a/test/registered/model_loading/test_load_weights_from_remote_instance.py +++ b/test/registered/model_loading/test_load_weights_from_remote_instance.py @@ -38,7 +38,7 @@ from sglang.utils import terminate_process mp.set_start_method("spawn", force=True) -register_cuda_ci(est_time=131, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=136, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=72, stage="extra-a", runner_config="2-gpu-large-amd") diff --git a/test/registered/model_loading/test_startup_weight_load.py b/test/registered/model_loading/test_startup_weight_load.py index cee4f1cec2..734d3815c9 100644 --- a/test/registered/model_loading/test_startup_weight_load.py +++ b/test/registered/model_loading/test_startup_weight_load.py @@ -9,7 +9,7 @@ from sglang.test.test_utils import ( CustomTestCase, ) -register_cuda_ci(est_time=63, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=74, stage="base-b", runner_config="1-gpu-small") class TestStartupWeightLoad(CustomTestCase): diff --git a/test/registered/model_loading/test_weight_cache_daemon.py b/test/registered/model_loading/test_weight_cache_daemon.py index 1592d29690..00b0cb27dd 100644 --- a/test/registered/model_loading/test_weight_cache_daemon.py +++ b/test/registered/model_loading/test_weight_cache_daemon.py @@ -29,7 +29,7 @@ DEFAULT_MODEL = "Qwen/Qwen3-0.6B" # IPC handoff is exercised on every PR. Since the CI runner executes the whole # file per suite, TestWeightCacheDaemonTP2 self-skips when fewer than 2 GPUs are # visible (i.e. on the 1-gpu runner). -register_cuda_ci(est_time=313, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=398, stage="extra-a", runner_config="2-gpu-large") register_cuda_ci(est_time=68, stage="base-b", runner_config="1-gpu-small") # Capture the client server's logs so test_loaded_via_ipc can assert the IPC diff --git a/test/registered/model_loading/test_weight_loader_v2_e2e.py b/test/registered/model_loading/test_weight_loader_v2_e2e.py index 005efe2ca0..699b23ab00 100644 --- a/test/registered/model_loading/test_weight_loader_v2_e2e.py +++ b/test/registered/model_loading/test_weight_loader_v2_e2e.py @@ -14,7 +14,7 @@ from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=62, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=81, stage="base-b", runner_config="1-gpu-small") import multiprocessing as mp diff --git a/test/registered/moe/test_cutedsl_moe.py b/test/registered/moe/test_cutedsl_moe.py index 48d471ea83..0f999608c4 100644 --- a/test/registered/moe/test_cutedsl_moe.py +++ b/test/registered/moe/test_cutedsl_moe.py @@ -21,7 +21,7 @@ except ImportError: CuteDslMoEWrapper = None convert_sf_to_mma_layout = None -register_cuda_ci(est_time=17, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=33, stage="extra-b", runner_config="4-gpu-b200") SKIP_TEST = torch.cuda.get_device_capability() < (10, 0) SKIP_REASON = "Nvfp4 Requires compute capability of 10 or above." diff --git a/test/registered/moe/test_hpc_ops_moe.py b/test/registered/moe/test_hpc_ops_moe.py index 40af9d0097..f749ea00cc 100644 --- a/test/registered/moe/test_hpc_ops_moe.py +++ b/test/registered/moe/test_hpc_ops_moe.py @@ -24,7 +24,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.layer_ut_utils import init_single_process_dist from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-large") # Qwen3-30B-A3B-FP8 MoE shapes. E, TOPK, H, I = 128, 8, 2048, 768 diff --git a/test/registered/moe/test_moe_ep_extra.py b/test/registered/moe/test_moe_ep_extra.py index a04e0f761b..983be7c911 100644 --- a/test/registered/moe/test_moe_ep_extra.py +++ b/test/registered/moe/test_moe_ep_extra.py @@ -18,7 +18,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=94, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=138, stage="extra-a", runner_config="2-gpu-large") class TestEpDeepGEMM(CustomTestCase): diff --git a/test/registered/moe/test_zero_experts.py b/test/registered/moe/test_zero_experts.py index 7576ae2a14..870b215985 100644 --- a/test/registered/moe/test_zero_experts.py +++ b/test/registered/moe/test_zero_experts.py @@ -6,7 +6,7 @@ from sglang.kernels.ops.moe.ep_moe_kernels import zero_experts_compute_triton from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=19, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=5, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/observability/test_customized_info_streaming.py b/test/registered/observability/test_customized_info_streaming.py index 0a81043765..55ae43121d 100644 --- a/test/registered/observability/test_customized_info_streaming.py +++ b/test/registered/observability/test_customized_info_streaming.py @@ -16,7 +16,7 @@ if TYPE_CHECKING: from sglang.srt.layers.logits_processor import LogitsProcessorOutput from sglang.srt.sampling.sampling_batch_info import SamplingBatchInfo -register_cuda_ci(est_time=26, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=34, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=120, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/observability/test_encoder_server_metrics.py b/test/registered/observability/test_encoder_server_metrics.py index 4a6bc2679e..73ba6017a8 100644 --- a/test/registered/observability/test_encoder_server_metrics.py +++ b/test/registered/observability/test_encoder_server_metrics.py @@ -23,7 +23,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=32, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=34, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=180, stage="stage-b", runner_config="1-gpu-small-amd") _MODEL_NAME = DEFAULT_SMALL_VLM_MODEL_NAME_FOR_TEST diff --git a/test/registered/observability/test_metrics.py b/test/registered/observability/test_metrics.py index 5105e54d40..7f50303a46 100644 --- a/test/registered/observability/test_metrics.py +++ b/test/registered/observability/test_metrics.py @@ -25,7 +25,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=144, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=182, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=32, suite="stage-b-test-1-gpu-small-amd") _MODEL_NAME = "Qwen/Qwen3-0.6B" diff --git a/test/registered/observability/test_tracing.py b/test/registered/observability/test_tracing.py index 3bda142419..ef497504b0 100644 --- a/test/registered/observability/test_tracing.py +++ b/test/registered/observability/test_tracing.py @@ -46,7 +46,7 @@ from sglang.test.test_utils import ( logger = logging.getLogger(__name__) # CI registration -register_cuda_ci(est_time=113, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=330, stage="extra-a", runner_config="1-gpu-small") # Backend-specific: the span assertions require PREFILL_FORWARD/DECODE_FORWARD # to be emitted from the scheduler forward path, which ROCm reaches through its # own attention backend and graph replay. diff --git a/test/registered/observability/test_tracing_disaggregation.py b/test/registered/observability/test_tracing_disaggregation.py index eeab8fc9f9..3fea8b3644 100644 --- a/test/registered/observability/test_tracing_disaggregation.py +++ b/test/registered/observability/test_tracing_disaggregation.py @@ -34,7 +34,7 @@ from sglang.utils import wait_for_http_ready logger = logging.getLogger(__name__) # CI registration - PD disaggregation requires 2 GPUs -register_cuda_ci(est_time=68, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=90, stage="base-b", runner_config="2-gpu-large") class TestTraceDisaggregation(CustomTestCase): diff --git a/test/registered/openai_server/basic/test_openai_completion_rust.py b/test/registered/openai_server/basic/test_openai_completion_rust.py index d39a1b14d5..443ec4a10c 100644 --- a/test/registered/openai_server/basic/test_openai_completion_rust.py +++ b/test/registered/openai_server/basic/test_openai_completion_rust.py @@ -14,7 +14,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=156, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=117, stage="base-b", runner_config="1-gpu-small") @unittest.skipUnless( diff --git a/test/registered/openai_server/basic/test_openai_server.py b/test/registered/openai_server/basic/test_openai_server.py index 09a550d856..aef51359c0 100644 --- a/test/registered/openai_server/basic/test_openai_server.py +++ b/test/registered/openai_server/basic/test_openai_server.py @@ -31,7 +31,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=353, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=420, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=280, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/openai_server/validation/test_matched_stop.py b/test/registered/openai_server/validation/test_matched_stop.py index 4846103f82..991cd756df 100644 --- a/test/registered/openai_server/validation/test_matched_stop.py +++ b/test/registered/openai_server/validation/test_matched_stop.py @@ -13,7 +13,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=52, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=82, stage="base-b", runner_config="1-gpu-small") register_cpu_ci(est_time=83, suite="stage-b-test-cpu-intel") diff --git a/test/registered/page_major/test_page_major_gpt_oss.py b/test/registered/page_major/test_page_major_gpt_oss.py index 1fea671c60..bf9cde3a59 100644 --- a/test/registered/page_major/test_page_major_gpt_oss.py +++ b/test/registered/page_major/test_page_major_gpt_oss.py @@ -14,7 +14,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.server_fixtures.default_fixture import DefaultServerBase from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST_MXFP4_WITH_MOE -register_cuda_ci(est_time=276, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=590, stage="extra-a", runner_config="1-gpu-large") _UNIFIED_COMMON_ARGS = [ "--enable-unified-memory", diff --git a/test/registered/page_major/test_page_major_qwen_hybrid.py b/test/registered/page_major/test_page_major_qwen_hybrid.py index 032edfcca3..b3dd0f60ed 100644 --- a/test/registered/page_major/test_page_major_qwen_hybrid.py +++ b/test/registered/page_major/test_page_major_qwen_hybrid.py @@ -15,7 +15,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.server_fixtures.default_fixture import DefaultServerBase from sglang.test.test_utils import DEFAULT_HYBRID_GDN_SMALL_MODEL_NAME_FOR_TEST -register_cuda_ci(est_time=261, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=395, stage="extra-a", runner_config="1-gpu-large") _UNIFIED_COMMON_ARGS = [ "--trust-remote-code", diff --git a/test/registered/pp/test_pp_parallel_compat.py b/test/registered/pp/test_pp_parallel_compat.py index ceb431b43e..cb5fe11d9d 100644 --- a/test/registered/pp/test_pp_parallel_compat.py +++ b/test/registered/pp/test_pp_parallel_compat.py @@ -11,7 +11,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=363, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=330, stage="extra-b", runner_config="4-gpu-h100") QWEN3_MOE_MODEL_PATH = "Qwen/Qwen3-30B-A3B-FP8" diff --git a/test/registered/pp/test_pp_single_node.py b/test/registered/pp/test_pp_single_node.py index d2b6f2e219..8ca9c9cb2d 100644 --- a/test/registered/pp/test_pp_single_node.py +++ b/test/registered/pp/test_pp_single_node.py @@ -31,7 +31,7 @@ from sglang.test.test_utils import ( run_bench_one_batch_server, ) -register_cuda_ci(est_time=295, stage="base-c", runner_config="4-gpu-h100") +register_cuda_ci(est_time=280, stage="base-c", runner_config="4-gpu-h100") register_amd_ci(est_time=500, suite="stage-c-test-4-gpu-amd") diff --git a/test/registered/pp/test_pp_single_node_extra.py b/test/registered/pp/test_pp_single_node_extra.py index 19f49f65a4..847fb74c24 100644 --- a/test/registered/pp/test_pp_single_node_extra.py +++ b/test/registered/pp/test_pp_single_node_extra.py @@ -25,7 +25,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=342, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=330, stage="extra-b", runner_config="4-gpu-h100") register_amd_ci(est_time=350, suite="stage-c-test-4-gpu-amd") diff --git a/test/registered/prefill_only/test_embedding_models.py b/test/registered/prefill_only/test_embedding_models.py index bd6d976ef2..e2d8fec5d3 100644 --- a/test/registered/prefill_only/test_embedding_models.py +++ b/test/registered/prefill_only/test_embedding_models.py @@ -35,7 +35,7 @@ register_amd_ci( suite="stage-b-test-1-gpu-small-amd", disabled="see https://github.com/sgl-project/sglang/issues/11127", ) -register_cuda_ci(est_time=154, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=162, stage="base-b", runner_config="1-gpu-small") MODEL_TO_CONFIG = { "Alibaba-NLP/gte-Qwen2-1.5B-instruct": (1, 1e-5), diff --git a/test/registered/prefill_only/test_openai_embedding.py b/test/registered/prefill_only/test_openai_embedding.py index 8388b3567c..2510e51c34 100644 --- a/test/registered/prefill_only/test_openai_embedding.py +++ b/test/registered/prefill_only/test_openai_embedding.py @@ -17,7 +17,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=94, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=116, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=141, suite="stage-b-test-1-gpu-small-amd") register_cpu_ci(est_time=91, suite="stage-b-test-cpu-intel") diff --git a/test/registered/prefill_only/test_score_api.py b/test/registered/prefill_only/test_score_api.py index 76d8f9873e..f021ef91f9 100644 --- a/test/registered/prefill_only/test_score_api.py +++ b/test/registered/prefill_only/test_score_api.py @@ -27,7 +27,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=94, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=114, stage="base-b", runner_config="1-gpu-small") _MODEL = os.environ.get("TEST_MODEL_NAME", DEFAULT_SMALL_MODEL_NAME_FOR_TEST) diff --git a/test/registered/radix_cache/test_int8_mamba_checkpoint_e2e.py b/test/registered/radix_cache/test_int8_mamba_checkpoint_e2e.py index 0680845136..8cc66e117b 100644 --- a/test/registered/radix_cache/test_int8_mamba_checkpoint_e2e.py +++ b/test/registered/radix_cache/test_int8_mamba_checkpoint_e2e.py @@ -38,7 +38,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=344, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=150, stage="extra-b", runner_config="4-gpu-h100") class TestInt8MambaCheckpointE2E(KLDivergenceMixin, DefaultServerBase): diff --git a/test/registered/radix_cache/test_mamba2_extra_buffer_kl.py b/test/registered/radix_cache/test_mamba2_extra_buffer_kl.py index 0fe222028d..82b8e558be 100644 --- a/test/registered/radix_cache/test_mamba2_extra_buffer_kl.py +++ b/test/registered/radix_cache/test_mamba2_extra_buffer_kl.py @@ -26,7 +26,7 @@ from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.kits.kl_divergence_kit import KLDivergenceMixin from sglang.test.server_fixtures.default_fixture import DefaultServerBase -register_cuda_ci(est_time=154, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=158, stage="extra-a", runner_config="1-gpu-large") class TestMamba2ExtraBufferKL(KLDivergenceMixin, DefaultServerBase): diff --git a/test/registered/radix_cache/test_radix_attention.py b/test/registered/radix_cache/test_radix_attention.py index 5136aa81d2..89ebd3b2e2 100644 --- a/test/registered/radix_cache/test_radix_attention.py +++ b/test/registered/radix_cache/test_radix_attention.py @@ -18,7 +18,7 @@ from sglang.test.test_utils import ( ) # RadixAttention server integration tests -register_cuda_ci(est_time=148, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=164, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=100, suite="stage-b-test-1-gpu-small-amd") register_cpu_ci(est_time=405, suite="stage-b-test-cpu-intel") diff --git a/test/registered/radix_cache/test_radix_cache_hit.py b/test/registered/radix_cache/test_radix_cache_hit.py index 6ce983a04c..946afa732a 100644 --- a/test/registered/radix_cache/test_radix_cache_hit.py +++ b/test/registered/radix_cache/test_radix_cache_hit.py @@ -15,7 +15,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=71, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=90, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=55, suite="stage-b-test-1-gpu-small-amd") register_cpu_ci(est_time=113, suite="stage-b-test-cpu-intel") diff --git a/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_dsv4.py b/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_dsv4.py index 20224cd0d4..07d7f1f4d9 100644 --- a/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_dsv4.py +++ b/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_dsv4.py @@ -20,7 +20,7 @@ DSV4_FLASH_MODEL = os.environ.get( ) DSV4_FLASH_LAUNCH_TIMEOUT = 3600 -register_cuda_ci(est_time=210, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=253, stage="extra-b", runner_config="4-gpu-h100") class TestDeepSeekV4FlashUnifiedCacheLinkerKL( diff --git a/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_glm52.py b/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_glm52.py index 50792565d4..042d63e3c0 100644 --- a/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_glm52.py +++ b/test/registered/radix_cache/unified_radix_tree/linker/test_unified_cache_linker_kl_glm52.py @@ -19,7 +19,7 @@ from sglang.test.test_utils import ( GLM52_MODEL = os.environ.get("SGLANG_LINKER_GLM52_MODEL", "zai-org/GLM-5.2-FP8") GLM52_LAUNCH_TIMEOUT = 3600 -register_cuda_ci(est_time=383, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=311, stage="extra-b", runner_config="8-gpu-h200") class TestGLM52UnifiedCacheLinkerKL(UnifiedRadixTreeTestMixin, CustomTestCase): diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_cp.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_cp.py index 2f85430472..0e2cff9532 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_cp.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_cp.py @@ -11,7 +11,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=346, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=232, stage="extra-b", runner_config="4-gpu-h100") QWEN3_32B_MODEL = "Qwen/Qwen3-32B" diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dcp.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dcp.py index 1d72c8612a..1e6940fac0 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dcp.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dcp.py @@ -24,7 +24,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=280, stage="extra-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=314, stage="extra-b", runner_config="4-gpu-b200") KIMI_LINEAR_MODEL = "moonshotai/Kimi-Linear-48B-A3B-Instruct" DCP_SIZE = 4 diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4.py index 557fefadc0..eef2f02b2a 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4.py @@ -25,7 +25,7 @@ DSV4_FLASH_MODEL = "sgl-project/DeepSeek-V4-Flash-FP8" DSV4_DSPARK_MODEL = "deepseek-ai/DeepSeek-V4-Flash-DSpark" DSV4_FLASH_LAUNCH_TIMEOUT = 3600 -register_cuda_ci(est_time=1865, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=2250, stage="extra-b", runner_config="4-gpu-h100") def _assert_dsv4_decode_cached_tokens(result, history_len, output_len, label): diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4_pp.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4_pp.py index 40b6893b58..7e5e309348 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4_pp.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_dsv4_pp.py @@ -4,7 +4,7 @@ import test_unified_radix_cache_kl_dsv4 as dsv4_kl from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=478, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=494, stage="extra-b", runner_config="8-gpu-h200") class TestUnifiedDeepSeekV4FlashHiCachePP4TP2( diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_glm52.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_glm52.py index 9a3591e7f9..99f4afc3be 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_glm52.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_glm52.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( GLM5_MODEL = "zai-org/GLM-5.2-FP8" GLM5_LAUNCH_TIMEOUT = 3600 -register_cuda_ci(est_time=340, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=370, stage="extra-b", runner_config="8-gpu-h200") class TestGLM5UnifiedRadixCacheL3Accuracy(AccuracyTwoPassMixin, CustomTestCase): diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_hybrid_bitexact.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_hybrid_bitexact.py index c7c5358ba8..fcc4481b4a 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_hybrid_bitexact.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_hybrid_bitexact.py @@ -73,7 +73,7 @@ from sglang.test.test_utils import ( unified_radix_tree_server_env, ) -register_cuda_ci(est_time=2300, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=1367, stage="extra-a", runner_config="1-gpu-large") _MODEL_PATH = os.environ.get("INKLING_TEST_MODEL_PATH", "thinkingmachines/Inkling") _MODEL_REVISION = os.environ.get("INKLING_TEST_MODEL_REVISION", "test") diff --git a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_mamba.py b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_mamba.py index 940a5a214b..fe44157668 100644 --- a/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_mamba.py +++ b/test/registered/radix_cache/unified_radix_tree/test_unified_radix_cache_kl_mamba.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( terminate_and_kill_process_tree, ) -register_cuda_ci(est_time=742, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=729, stage="extra-b", runner_config="4-gpu-h100") MAMBA_MODEL = "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8" MAMBA_CHUNK_SIZE = 64 diff --git a/test/registered/rl/test_fp32_lm_head.py b/test/registered/rl/test_fp32_lm_head.py index 2e0f1676b0..69f5856afe 100644 --- a/test/registered/rl/test_fp32_lm_head.py +++ b/test/registered/rl/test_fp32_lm_head.py @@ -12,7 +12,7 @@ from sglang.srt.utils import get_device from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=28, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=15, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/rl/test_multi_instance_release_memory_occupation.py b/test/registered/rl/test_multi_instance_release_memory_occupation.py index fcb3b25235..e58b762afb 100644 --- a/test/registered/rl/test_multi_instance_release_memory_occupation.py +++ b/test/registered/rl/test_multi_instance_release_memory_occupation.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( find_available_port, ) -register_cuda_ci(est_time=69, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=62, stage="extra-b", runner_config="4-gpu-h100") register_amd_ci( est_time=64, suite="stage-c-test-4-gpu-amd", diff --git a/test/registered/rl/test_patch_torch.py b/test/registered/rl/test_patch_torch.py index 58fb805540..db622d4425 100644 --- a/test/registered/rl/test_patch_torch.py +++ b/test/registered/rl/test_patch_torch.py @@ -9,7 +9,7 @@ import torch.multiprocessing as mp from sglang.srt.utils.patch_torch import monkey_patch_torch_reductions from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=13, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=5, stage="base-b", runner_config="2-gpu-large") register_amd_ci(est_time=30, suite="stage-b-test-2-gpu-large-amd") diff --git a/test/registered/rl/test_return_indexer_topk.py b/test/registered/rl/test_return_indexer_topk.py index eb0c24096e..ab28eee826 100644 --- a/test/registered/rl/test_return_indexer_topk.py +++ b/test/registered/rl/test_return_indexer_topk.py @@ -17,7 +17,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=306, stage="extra-b", runner_config="8-gpu-h200") +register_cuda_ci(est_time=273, stage="extra-b", runner_config="8-gpu-h200") DEEPSEEK_V32_MODEL_PATH = "deepseek-ai/DeepSeek-V3.2" diff --git a/test/registered/rl/test_return_routed_experts.py b/test/registered/rl/test_return_routed_experts.py index fc54a43f32..3cd2840645 100644 --- a/test/registered/rl/test_return_routed_experts.py +++ b/test/registered/rl/test_return_routed_experts.py @@ -24,7 +24,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=446, stage="extra-b", runner_config="4-gpu-h100") +register_cuda_ci(est_time=479, stage="extra-b", runner_config="4-gpu-h100") # FP8 variant of Qwen3-30B-A3B: required because DeepEP normal/LL fast paths in # ep_moe/layer.py only run for {Fp8Config (via deep_gemm), W4AFp8Config, aiter, diff --git a/test/registered/rl/test_update_weights_from_distributed.py b/test/registered/rl/test_update_weights_from_distributed.py index ec47a36885..3625251d67 100644 --- a/test/registered/rl/test_update_weights_from_distributed.py +++ b/test/registered/rl/test_update_weights_from_distributed.py @@ -43,7 +43,7 @@ from sglang.test.test_utils import ( ) from sglang.utils import terminate_process -register_cuda_ci(est_time=120, stage="extra-a", runner_config="2-gpu-large") +register_cuda_ci(est_time=121, stage="extra-a", runner_config="2-gpu-large") register_amd_ci(est_time=400, suite="stage-b-test-2-gpu-large-amd") mp.set_start_method("spawn", force=True) diff --git a/test/registered/rl/test_update_weights_from_tensor.py b/test/registered/rl/test_update_weights_from_tensor.py index d9dd41a604..fee4210762 100644 --- a/test/registered/rl/test_update_weights_from_tensor.py +++ b/test/registered/rl/test_update_weights_from_tensor.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=165, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=197, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=195, suite="stage-b-test-1-gpu-small-amd") import gc diff --git a/test/registered/rotary/test_mrope_axis_map.py b/test/registered/rotary/test_mrope_axis_map.py index 3d60aac50f..88a8243729 100644 --- a/test/registered/rotary/test_mrope_axis_map.py +++ b/test/registered/rotary/test_mrope_axis_map.py @@ -12,7 +12,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def build_mrope( diff --git a/test/registered/rotary/test_rope_cache_invalidation.py b/test/registered/rotary/test_rope_cache_invalidation.py index be2fbdc5fb..42ce6454e0 100644 --- a/test/registered/rotary/test_rope_cache_invalidation.py +++ b/test/registered/rotary/test_rope_cache_invalidation.py @@ -9,7 +9,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") _ROPE_KWARGS = dict( head_size=64, diff --git a/test/registered/rust/test_rust_extension.py b/test/registered/rust/test_rust_extension.py index 5320813791..f14d733ffc 100644 --- a/test/registered/rust/test_rust_extension.py +++ b/test/registered/rust/test_rust_extension.py @@ -18,7 +18,7 @@ from sglang.srt.rust_extensions.torch_build import torch_build_configuration from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _hold_filesystem_lock(path: str, ready, release) -> None: diff --git a/test/registered/sampling/test_deterministic_gumbel_u1.py b/test/registered/sampling/test_deterministic_gumbel_u1.py index 5457574e1c..041191854e 100644 --- a/test/registered/sampling/test_deterministic_gumbel_u1.py +++ b/test/registered/sampling/test_deterministic_gumbel_u1.py @@ -9,7 +9,7 @@ from sglang.srt.layers.sampler import sampling_from_probs_torch from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=60, suite="stage-b-test-1-gpu-small-amd") VOCAB = 248320 diff --git a/test/registered/sampling/test_original_logprobs.py b/test/registered/sampling/test_original_logprobs.py index 029a8c7849..d34a8acb97 100644 --- a/test/registered/sampling/test_original_logprobs.py +++ b/test/registered/sampling/test_original_logprobs.py @@ -25,7 +25,7 @@ import sglang as sgl from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST -register_cuda_ci(est_time=52, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=85, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=60, suite="stage-b-test-1-gpu-small-amd") # ------------------------- Configurable via env ------------------------- # diff --git a/test/registered/sampling/test_penalty.py b/test/registered/sampling/test_penalty.py index 8f57658402..699e07b64a 100644 --- a/test/registered/sampling/test_penalty.py +++ b/test/registered/sampling/test_penalty.py @@ -16,7 +16,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=61, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=80, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=82, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/scheduler/test_abort_with_metrics.py b/test/registered/scheduler/test_abort_with_metrics.py index 5e955eafac..217be3f76d 100644 --- a/test/registered/scheduler/test_abort_with_metrics.py +++ b/test/registered/scheduler/test_abort_with_metrics.py @@ -18,7 +18,7 @@ from sglang.srt.utils.http_middleware_patch import _PureASGIDispatch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") _HTTP_SCOPE = { diff --git a/test/registered/scheduler/test_min_free_slots_delayer.py b/test/registered/scheduler/test_min_free_slots_delayer.py index 1e573bee44..8793083dba 100644 --- a/test/registered/scheduler/test_min_free_slots_delayer.py +++ b/test/registered/scheduler/test_min_free_slots_delayer.py @@ -6,7 +6,7 @@ from sglang.srt.managers.min_free_slots_delayer import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestResolveMinFreeSlots(unittest.TestCase): diff --git a/test/registered/scheduler/test_mixed_chunked_prefill.py b/test/registered/scheduler/test_mixed_chunked_prefill.py index 3a944bca4b..0f31d21c30 100644 --- a/test/registered/scheduler/test_mixed_chunked_prefill.py +++ b/test/registered/scheduler/test_mixed_chunked_prefill.py @@ -12,7 +12,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=176, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=194, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=180, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/scheduler/test_priority_scheduling.py b/test/registered/scheduler/test_priority_scheduling.py index bbd2b13a82..942ddcdf16 100644 --- a/test/registered/scheduler/test_priority_scheduling.py +++ b/test/registered/scheduler/test_priority_scheduling.py @@ -17,7 +17,7 @@ from sglang.test.test_utils import ( send_concurrent_generate_requests_with_custom_params, ) -register_cuda_ci(est_time=166, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=168, stage="extra-a", runner_config="1-gpu-small") register_amd_ci(est_time=195, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/scheduler/test_retract_decode_logprob.py b/test/registered/scheduler/test_retract_decode_logprob.py index 0e63cccf98..c1dd5cd035 100644 --- a/test/registered/scheduler/test_retract_decode_logprob.py +++ b/test/registered/scheduler/test_retract_decode_logprob.py @@ -56,7 +56,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=76, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=93, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=360, suite="stage-b-test-1-gpu-small-amd") N_REQUESTS = 32 diff --git a/test/registered/scheduler/test_scheduler_control.py b/test/registered/scheduler/test_scheduler_control.py index 4ac9ec28a3..64db2088a3 100644 --- a/test/registered/scheduler/test_scheduler_control.py +++ b/test/registered/scheduler/test_scheduler_control.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import ( run_and_check_memory_leak, ) -register_cuda_ci(est_time=421, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=512, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=300, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/scripted_runtime/test_scripted_runtime_core.py b/test/registered/scripted_runtime/test_scripted_runtime_core.py index d6829552ea..80aa6240b6 100644 --- a/test/registered/scripted_runtime/test_scripted_runtime_core.py +++ b/test/registered/scripted_runtime/test_scripted_runtime_core.py @@ -16,7 +16,7 @@ from sglang.test.scripted_runtime_chunked_helpers import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=297, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=339, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=460, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/sessions/test_session_control.py b/test/registered/sessions/test_session_control.py index fbe2498d40..e597e541b5 100644 --- a/test/registered/sessions/test_session_control.py +++ b/test/registered/sessions/test_session_control.py @@ -28,7 +28,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=137, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=220, stage="extra-a", runner_config="1-gpu-large") register_amd_ci(est_time=87, suite="stage-b-test-1-gpu-large-amd") diff --git a/test/registered/spec/dspark/test_dspark_block_accept_estimator.py b/test/registered/spec/dspark/test_dspark_block_accept_estimator.py index 583fba4228..6dfbd9f439 100644 --- a/test/registered/spec/dspark/test_dspark_block_accept_estimator.py +++ b/test/registered/spec/dspark/test_dspark_block_accept_estimator.py @@ -12,7 +12,7 @@ from sglang.srt.speculative.dspark_components.dspark_block_accept_estimator impo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _GAMMA = 3 _VOCAB = 8 diff --git a/test/registered/spec/dspark/test_dspark_confidence_metrics.py b/test/registered/spec/dspark/test_dspark_confidence_metrics.py index 61d7436bc1..ce1f317f9e 100644 --- a/test/registered/spec/dspark/test_dspark_confidence_metrics.py +++ b/test/registered/spec/dspark/test_dspark_confidence_metrics.py @@ -10,7 +10,7 @@ from sglang.srt.speculative.dspark_components.dspark_observability import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _cpu_metrics(gamma: int) -> PerPositionConfidenceMetrics: diff --git a/test/registered/spec/dspark/test_dspark_dp_tier.py b/test/registered/spec/dspark/test_dspark_dp_tier.py index 95f3a1ca88..b8922ae36c 100644 --- a/test/registered/spec/dspark/test_dspark_dp_tier.py +++ b/test/registered/spec/dspark/test_dspark_dp_tier.py @@ -13,7 +13,7 @@ from sglang.srt.speculative.dspark_components.dspark_planner import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestLocalVerifyTierNumTokens(CustomTestCase): diff --git a/test/registered/spec/dspark/test_dspark_draft_path_default.py b/test/registered/spec/dspark/test_dspark_draft_path_default.py index b0da41feb8..0a907d2d4f 100644 --- a/test/registered/spec/dspark/test_dspark_draft_path_default.py +++ b/test/registered/spec/dspark/test_dspark_draft_path_default.py @@ -13,7 +13,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") _BUNDLED_MODEL_PATH = "deepseek-ai/DeepSeek-V4-Flash-DSpark" _PLAIN_MODEL_PATH = "deepseek-ai/DeepSeek-V4-Flash" diff --git a/test/registered/spec/dspark/test_dspark_info_dumper.py b/test/registered/spec/dspark/test_dspark_info_dumper.py index abd328735d..dc0d975d71 100644 --- a/test/registered/spec/dspark/test_dspark_info_dumper.py +++ b/test/registered/spec/dspark/test_dspark_info_dumper.py @@ -22,7 +22,7 @@ from sglang.srt.utils.msgspec_utils import msgspec_to_builtins from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class FakeClock: diff --git a/test/registered/spec/dspark/test_dspark_kernel_parity.py b/test/registered/spec/dspark/test_dspark_kernel_parity.py index ed583d08af..7340d30212 100644 --- a/test/registered/spec/dspark/test_dspark_kernel_parity.py +++ b/test/registered/spec/dspark/test_dspark_kernel_parity.py @@ -30,7 +30,7 @@ from sglang.srt.speculative.ragged_verify import RaggedVerifyLayout from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=26, stage="base-b", runner_config="1-gpu-small") DEVICE = torch.device("cuda") VOCAB = 129280 diff --git a/test/registered/spec/dspark/test_dspark_scheduler.py b/test/registered/spec/dspark/test_dspark_scheduler.py index bc1a030a67..977375f4ac 100644 --- a/test/registered/spec/dspark/test_dspark_scheduler.py +++ b/test/registered/spec/dspark/test_dspark_scheduler.py @@ -22,7 +22,7 @@ from sglang.srt.speculative.ragged_verify import RaggedVerifyLayout from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _flat_table( diff --git a/test/registered/spec/dspark/test_dspark_sps_table.py b/test/registered/spec/dspark/test_dspark_sps_table.py index bdab10159e..9973d98e4f 100644 --- a/test/registered/spec/dspark/test_dspark_sps_table.py +++ b/test/registered/spec/dspark/test_dspark_sps_table.py @@ -13,7 +13,7 @@ from sglang.srt.speculative.dspark_components.dspark_sps import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_table() -> SpsCostTable: diff --git a/test/registered/spec/dspark/test_dspark_stacked_ctx_kv_parity.py b/test/registered/spec/dspark/test_dspark_stacked_ctx_kv_parity.py index c3891d027a..539f1116cf 100644 --- a/test/registered/spec/dspark/test_dspark_stacked_ctx_kv_parity.py +++ b/test/registered/spec/dspark/test_dspark_stacked_ctx_kv_parity.py @@ -19,7 +19,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=22, stage="base-b", runner_config="1-gpu-small") DEVICE = torch.device("cuda") HEAD_DIM = 64 diff --git a/test/registered/spec/dspark/test_dspark_sts.py b/test/registered/spec/dspark/test_dspark_sts.py index b7b890c65b..45c1243858 100644 --- a/test/registered/spec/dspark/test_dspark_sts.py +++ b/test/registered/spec/dspark/test_dspark_sts.py @@ -17,7 +17,7 @@ from sglang.srt.speculative.dspark_components.dspark_sts import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestApplySts(CustomTestCase): diff --git a/test/registered/spec/dspark/test_ragged_verify.py b/test/registered/spec/dspark/test_ragged_verify.py index 95a713b055..6690e986f0 100644 --- a/test/registered/spec/dspark/test_ragged_verify.py +++ b/test/registered/spec/dspark/test_ragged_verify.py @@ -9,7 +9,7 @@ from sglang.srt.speculative.ragged_verify import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") _DEVICE = torch.device("cpu") _GRID = [8, 16, 24, 32, 64] diff --git a/test/registered/spec/dspark/test_ragged_verify_backend_capability.py b/test/registered/spec/dspark/test_ragged_verify_backend_capability.py index b0257ae28b..a776b17889 100644 --- a/test/registered/spec/dspark/test_ragged_verify_backend_capability.py +++ b/test/registered/spec/dspark/test_ragged_verify_backend_capability.py @@ -9,7 +9,7 @@ import unittest from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small") class TestRaggedVerifyGraphCapability(CustomTestCase): diff --git a/test/registered/spec/eagle/test_eagle_dp_attention.py b/test/registered/spec/eagle/test_eagle_dp_attention.py index 111429a6fe..db6923733d 100644 --- a/test/registered/spec/eagle/test_eagle_dp_attention.py +++ b/test/registered/spec/eagle/test_eagle_dp_attention.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( ) # EAGLE3 with DP attention (tp=2, dp=2, requires 4 GPUs). -register_cuda_ci(est_time=112, stage="base-c", runner_config="4-gpu-h100") +register_cuda_ci(est_time=103, stage="base-c", runner_config="4-gpu-h100") register_amd_ci(est_time=200, suite="stage-c-test-4-gpu-amd") diff --git a/test/registered/spec/eagle/test_eagle_reject_sampling.py b/test/registered/spec/eagle/test_eagle_reject_sampling.py index feb82f333c..aa55ef29e7 100644 --- a/test/registered/spec/eagle/test_eagle_reject_sampling.py +++ b/test/registered/spec/eagle/test_eagle_reject_sampling.py @@ -15,7 +15,7 @@ from sglang.test.test_utils import ( write_github_step_summary, ) -register_cuda_ci(est_time=81, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=94, stage="base-b", runner_config="2-gpu-large") QWEN35_MODEL = "Qwen/Qwen3.5-9B" SERVER_LAUNCH_TIMEOUT = 600 diff --git a/test/registered/spec/eagle/test_spec_eagle.py b/test/registered/spec/eagle/test_spec_eagle.py index 3a4337215b..ba06044732 100644 --- a/test/registered/spec/eagle/test_spec_eagle.py +++ b/test/registered/spec/eagle/test_spec_eagle.py @@ -20,7 +20,7 @@ from sglang.test.kits.spec_server_kits import ( ) from sglang.test.server_fixtures.spec_eagle_fixture import Eagle3Base -register_cuda_ci(est_time=403, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=416, stage="base-b", runner_config="1-gpu-small") _KITS = ( SpecCorrectnessKit, diff --git a/test/registered/spec/eagle/test_spec_eagle_fa3.py b/test/registered/spec/eagle/test_spec_eagle_fa3.py index 6196ff72eb..ef13b8a8d9 100644 --- a/test/registered/spec/eagle/test_spec_eagle_fa3.py +++ b/test/registered/spec/eagle/test_spec_eagle_fa3.py @@ -17,7 +17,7 @@ from sglang.test.kits.spec_server_kits import ( ) from sglang.test.server_fixtures.spec_eagle_fixture import Eagle3Base, EagleLlama2Base -register_cuda_ci(est_time=281, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=437, stage="base-b", runner_config="1-gpu-large") class TestEagle3Fa3(Eagle3Base, SpecAccuracyKit, SpecLogprobKit): diff --git a/test/registered/spec/eagle/test_spec_eagle_stress.py b/test/registered/spec/eagle/test_spec_eagle_stress.py index a6b229cae2..53b6456076 100644 --- a/test/registered/spec/eagle/test_spec_eagle_stress.py +++ b/test/registered/spec/eagle/test_spec_eagle_stress.py @@ -17,7 +17,7 @@ from sglang.test.kits.spec_server_kits import ( ) from sglang.test.server_fixtures.spec_eagle_fixture import Eagle3Base, EagleLlama2Base -register_cuda_ci(est_time=440, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=568, stage="extra-a", runner_config="1-gpu-large") class TestEagle3Perf(Eagle3Base, SpecPerfKit): diff --git a/test/registered/spec/eagle/test_spec_eagle_triton.py b/test/registered/spec/eagle/test_spec_eagle_triton.py index 1143fd1b83..fd1603bae9 100644 --- a/test/registered/spec/eagle/test_spec_eagle_triton.py +++ b/test/registered/spec/eagle/test_spec_eagle_triton.py @@ -18,7 +18,7 @@ from sglang.test.kits.spec_server_kits import ( ) from sglang.test.server_fixtures.spec_eagle_fixture import Eagle3Base -register_cuda_ci(est_time=191, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=189, stage="base-b", runner_config="1-gpu-small") class TestEagle3Triton( diff --git a/test/registered/spec/test_spec_mixed_chunk.py b/test/registered/spec/test_spec_mixed_chunk.py index 8cd1b09165..1e2da75024 100644 --- a/test/registered/spec/test_spec_mixed_chunk.py +++ b/test/registered/spec/test_spec_mixed_chunk.py @@ -32,7 +32,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=196, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=339, stage="base-b", runner_config="1-gpu-large") class TestEagle3MixedChunk( diff --git a/test/registered/spec/test_spec_ngram.py b/test/registered/spec/test_spec_ngram.py index f388d0645d..d729496e05 100644 --- a/test/registered/spec/test_spec_ngram.py +++ b/test/registered/spec/test_spec_ngram.py @@ -10,7 +10,7 @@ from sglang.test.server_fixtures.ngram_fixture import NgramServerBase # Per-commit: Paged backend only. # - FA3 base test archived to test/manual/spec/test_spec_ngram_fa3.py # - Triton + Flashinfer moved to test_spec_ngram_extra.py -register_cuda_ci(est_time=82, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=123, stage="base-b", runner_config="1-gpu-large") class TestNgramSpeculativeDecodingPaged( diff --git a/test/registered/spec/test_spec_standalone.py b/test/registered/spec/test_spec_standalone.py index 96d171b214..6fcee8c9fa 100644 --- a/test/registered/spec/test_spec_standalone.py +++ b/test/registered/spec/test_spec_standalone.py @@ -10,7 +10,7 @@ from sglang.test.test_utils import CustomTestCase, is_in_ci # deployed on); triton / flashinfer stay runnable locally, and their spec verify # numerics live in attention/unittests/dense/test_{triton,flashinfer}.py. # Non-V2 backends moved to test_spec_standalone_extra.py. -register_cuda_ci(est_time=74, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=129, stage="base-b", runner_config="1-gpu-large") class TestStandaloneV2SpeculativeDecodingBase( diff --git a/test/registered/spec/test_spec_standalone_extra.py b/test/registered/spec/test_spec_standalone_extra.py index c43dbeb023..d833505e21 100644 --- a/test/registered/spec/test_spec_standalone_extra.py +++ b/test/registered/spec/test_spec_standalone_extra.py @@ -7,7 +7,7 @@ from sglang.test.test_utils import CustomTestCase # Non-V2 standalone speculative decoding tests (FA3, Triton, FlashInfer # backends). Sibling V2 classes stay per-commit in test_spec_standalone.py. -register_cuda_ci(est_time=225, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=344, stage="extra-a", runner_config="1-gpu-large") # AMD: fa3 / flashinfer attention backends are not built in the ROCm # sgl_kernel, so only the triton-backend class runs on ROCm (the fa3 and # flashinfer classes are skipped on ROCm below). diff --git a/test/registered/spec/utils/test_build_eagle_tree.py b/test/registered/spec/utils/test_build_eagle_tree.py index d2be6496bb..25b2e40a06 100644 --- a/test/registered/spec/utils/test_build_eagle_tree.py +++ b/test/registered/spec/utils/test_build_eagle_tree.py @@ -9,7 +9,7 @@ from sglang.srt.speculative.eagle_utils import ( from sglang.srt.utils import get_device from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=4, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/unit/batch_overlap/test_tbo_children_dummy_token_mask.py b/test/registered/unit/batch_overlap/test_tbo_children_dummy_token_mask.py index 9430a5faac..b1a9c7e7ba 100644 --- a/test/registered/unit/batch_overlap/test_tbo_children_dummy_token_mask.py +++ b/test/registered/unit/batch_overlap/test_tbo_children_dummy_token_mask.py @@ -26,7 +26,7 @@ from sglang.srt.runtime_context import get_context, get_device, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") def _make_extend_batch(*, padded_num_tokens: int, global_num_token_non_padded_cpu: int): diff --git a/test/registered/unit/batch_overlap/test_tbo_cuda_graph_num_token_device.py b/test/registered/unit/batch_overlap/test_tbo_cuda_graph_num_token_device.py index fc3fc75302..5067fcdeab 100644 --- a/test/registered/unit/batch_overlap/test_tbo_cuda_graph_num_token_device.py +++ b/test/registered/unit/batch_overlap/test_tbo_cuda_graph_num_token_device.py @@ -29,7 +29,7 @@ from sglang.srt.runtime_context import get_context, get_device from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestTboCudaGraphNumTokenDevice(CustomTestCase): diff --git a/test/registered/unit/batch_overlap/test_tbo_filter_batch_marker.py b/test/registered/unit/batch_overlap/test_tbo_filter_batch_marker.py index cba5a44efd..11e3c9fcfe 100644 --- a/test/registered/unit/batch_overlap/test_tbo_filter_batch_marker.py +++ b/test/registered/unit/batch_overlap/test_tbo_filter_batch_marker.py @@ -15,7 +15,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_target_verify_batch(bs: int) -> ForwardBatch: diff --git a/test/registered/unit/beam_search/test_beam_search_core.py b/test/registered/unit/beam_search/test_beam_search_core.py index c22fce99fb..68df4cce36 100644 --- a/test/registered/unit/beam_search/test_beam_search_core.py +++ b/test/registered/unit/beam_search/test_beam_search_core.py @@ -20,7 +20,7 @@ from sglang.srt.beam_search import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def T(data, dtype): diff --git a/test/registered/unit/beam_search/test_fork.py b/test/registered/unit/beam_search/test_fork.py index a199b61556..872894137b 100644 --- a/test/registered/unit/beam_search/test_fork.py +++ b/test/registered/unit/beam_search/test_fork.py @@ -18,7 +18,7 @@ from sglang.srt.managers.schedule_batch import ReqKvInfo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestRemapKvMapping(CustomTestCase): diff --git a/test/registered/unit/beam_search/test_output_decode.py b/test/registered/unit/beam_search/test_output_decode.py index 21a8f0e596..1febce091b 100644 --- a/test/registered/unit/beam_search/test_output_decode.py +++ b/test/registered/unit/beam_search/test_output_decode.py @@ -19,7 +19,7 @@ from sglang.srt.managers.detokenizer_manager import DetokenizerManager from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") STOP_ID = 99 diff --git a/test/registered/unit/bench/test_benchmark_endpoint.py b/test/registered/unit/bench/test_benchmark_endpoint.py index 61e39899bf..31384859b5 100644 --- a/test/registered/unit/bench/test_benchmark_endpoint.py +++ b/test/registered/unit/bench/test_benchmark_endpoint.py @@ -8,7 +8,7 @@ from sglang.benchmark.endpoint import launch_or_reuse_server from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _noop_launch_server(server_args): diff --git a/test/registered/unit/cli/test_serve_backends.py b/test/registered/unit/cli/test_serve_backends.py index f76cb089b8..aeaaad73c0 100644 --- a/test/registered/unit/cli/test_serve_backends.py +++ b/test/registered/unit/cli/test_serve_backends.py @@ -14,7 +14,7 @@ from sglang.cli.serve_backends import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _make_entry_point(name, factory, distribution=None): diff --git a/test/registered/unit/compilation/test_torch_compile_decoration.py b/test/registered/unit/compilation/test_torch_compile_decoration.py index 7d94466660..82d7b24647 100644 --- a/test/registered/unit/compilation/test_torch_compile_decoration.py +++ b/test/registered/unit/compilation/test_torch_compile_decoration.py @@ -6,7 +6,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") from sglang.srt.compilation.compile import ( _infer_dynamic_arg_dims_from_annotations, diff --git a/test/registered/unit/configs/test_cohere2_moe_config.py b/test/registered/unit/configs/test_cohere2_moe_config.py index 086674e8a7..eda17a2fa0 100644 --- a/test/registered/unit/configs/test_cohere2_moe_config.py +++ b/test/registered/unit/configs/test_cohere2_moe_config.py @@ -12,7 +12,7 @@ import unittest from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class TestCohere2MoeConfig(CustomTestCase): diff --git a/test/registered/unit/configs/test_embedding_model_spec.py b/test/registered/unit/configs/test_embedding_model_spec.py index 23dff18784..456f958651 100644 --- a/test/registered/unit/configs/test_embedding_model_spec.py +++ b/test/registered/unit/configs/test_embedding_model_spec.py @@ -14,7 +14,7 @@ from sglang.srt.configs.embedding_model_spec import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestEmbeddingModelSpec(unittest.TestCase): diff --git a/test/registered/unit/configs/test_laguna_config.py b/test/registered/unit/configs/test_laguna_config.py index 3e39b59879..82c9aa3930 100644 --- a/test/registered/unit/configs/test_laguna_config.py +++ b/test/registered/unit/configs/test_laguna_config.py @@ -19,7 +19,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # (factor, attention_factor). S/XS ship attention_factor == the yarn default for # their factor; _NON_DEFAULT differs from both 1.0 and the default, so a naive diff --git a/test/registered/unit/configs/test_linear_attn_model_registry.py b/test/registered/unit/configs/test_linear_attn_model_registry.py index 82c662c66c..83fccc5874 100644 --- a/test/registered/unit/configs/test_linear_attn_model_registry.py +++ b/test/registered/unit/configs/test_linear_attn_model_registry.py @@ -13,7 +13,7 @@ from sglang.srt.configs.linear_attn_model_registry import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") # Dummy config classes for testing diff --git a/test/registered/unit/configs/test_locate_anything_config.py b/test/registered/unit/configs/test_locate_anything_config.py index 97ea6f7004..4caa7e152e 100644 --- a/test/registered/unit/configs/test_locate_anything_config.py +++ b/test/registered/unit/configs/test_locate_anything_config.py @@ -9,7 +9,7 @@ from sglang.srt.configs.kimi_vl_moonvit import MoonViTConfig from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestLocateAnythingConfig(CustomTestCase): diff --git a/test/registered/unit/configs/test_minicpm_config.py b/test/registered/unit/configs/test_minicpm_config.py index e97f3c3b88..97bd86ea65 100644 --- a/test/registered/unit/configs/test_minicpm_config.py +++ b/test/registered/unit/configs/test_minicpm_config.py @@ -29,7 +29,7 @@ from sglang.srt.models.minicpm import ( from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def test_minicpm_lightning_config_defaults_are_complete(): diff --git a/test/registered/unit/configs/test_model_config.py b/test/registered/unit/configs/test_model_config.py index 73edb2c34e..e0cf06354d 100644 --- a/test/registered/unit/configs/test_model_config.py +++ b/test/registered/unit/configs/test_model_config.py @@ -21,7 +21,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestHybridLayerIds(CustomTestCase): diff --git a/test/registered/unit/configs/test_model_config_parser_registry.py b/test/registered/unit/configs/test_model_config_parser_registry.py index c647814262..0821437d8c 100644 --- a/test/registered/unit/configs/test_model_config_parser_registry.py +++ b/test/registered/unit/configs/test_model_config_parser_registry.py @@ -13,7 +13,7 @@ from sglang.srt.configs.model_config_parser_registry import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class _FakeParser(ModelConfigParserBase): diff --git a/test/registered/unit/configs/test_model_config_scaling.py b/test/registered/unit/configs/test_model_config_scaling.py index 6483890d2c..704e7a5103 100644 --- a/test/registered/unit/configs/test_model_config_scaling.py +++ b/test/registered/unit/configs/test_model_config_scaling.py @@ -6,7 +6,7 @@ from sglang.srt.configs.model_config import ModelConfig, compute_mla_mscale_scal from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _mla_scaling(rope_scaling) -> tuple[float, float]: diff --git a/test/registered/unit/configs/test_model_config_shapes.py b/test/registered/unit/configs/test_model_config_shapes.py index 651dc87d73..972e9dd2c9 100644 --- a/test/registered/unit/configs/test_model_config_shapes.py +++ b/test/registered/unit/configs/test_model_config_shapes.py @@ -10,7 +10,7 @@ from sglang.srt.configs.model_config import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") def _make_text_config(**overrides): diff --git a/test/registered/unit/configs/test_multimodal_piecewise_cuda_graph.py b/test/registered/unit/configs/test_multimodal_piecewise_cuda_graph.py index b76e82bb70..b55f158f98 100644 --- a/test/registered/unit/configs/test_multimodal_piecewise_cuda_graph.py +++ b/test/registered/unit/configs/test_multimodal_piecewise_cuda_graph.py @@ -30,7 +30,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestMultimodalPiecewiseCudaGraph(CustomTestCase): diff --git a/test/registered/unit/configs/test_nano_nemotron_vl_config.py b/test/registered/unit/configs/test_nano_nemotron_vl_config.py index a8612ccb28..e0a90705fd 100644 --- a/test/registered/unit/configs/test_nano_nemotron_vl_config.py +++ b/test/registered/unit/configs/test_nano_nemotron_vl_config.py @@ -8,7 +8,7 @@ from sglang.srt.configs.nano_nemotron_vl import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestNemotronHOmniConfig(CustomTestCase): diff --git a/test/registered/unit/configs/test_zaya_config.py b/test/registered/unit/configs/test_zaya_config.py index 9642b889fb..48ac0f02c2 100644 --- a/test/registered/unit/configs/test_zaya_config.py +++ b/test/registered/unit/configs/test_zaya_config.py @@ -8,7 +8,7 @@ from sglang.srt.configs.zaya import ZayaConfig, register_zaya_config from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestZayaConfig(CustomTestCase): diff --git a/test/registered/unit/constrained/test_llguidance_batched_mask.py b/test/registered/unit/constrained/test_llguidance_batched_mask.py index 5737aa3538..ab2c6b1088 100644 --- a/test/registered/unit/constrained/test_llguidance_batched_mask.py +++ b/test/registered/unit/constrained/test_llguidance_batched_mask.py @@ -11,7 +11,7 @@ from sglang.srt.constrained.llguidance_backend import GuidanceBackend, GuidanceG from sglang.srt.runtime_context import get_resources from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") _REGEX = r"[0-9]{1,8}" diff --git a/test/registered/unit/disaggregation/test_admission_abort_not_enqueued.py b/test/registered/unit/disaggregation/test_admission_abort_not_enqueued.py index e4ef48bf9c..0040424b4d 100644 --- a/test/registered/unit/disaggregation/test_admission_abort_not_enqueued.py +++ b/test/registered/unit/disaggregation/test_admission_abort_not_enqueued.py @@ -36,7 +36,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") ERROR_MSG = ( "Input length (1500 tokens) exceeds the maximum allowed length (1018 tokens)." diff --git a/test/registered/unit/disaggregation/test_dcp_pack.py b/test/registered/unit/disaggregation/test_dcp_pack.py index dc5190d372..0caf4e5746 100644 --- a/test/registered/unit/disaggregation/test_dcp_pack.py +++ b/test/registered/unit/disaggregation/test_dcp_pack.py @@ -21,7 +21,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") def _plan(*, src, dst, page_size, dcp_size, dcp_rank, **kwargs): diff --git a/test/registered/unit/disaggregation/test_decode_hicache_tree_core.py b/test/registered/unit/disaggregation/test_decode_hicache_tree_core.py index adc255ed65..f0a6e1dabd 100644 --- a/test/registered/unit/disaggregation/test_decode_hicache_tree_core.py +++ b/test/registered/unit/disaggregation/test_decode_hicache_tree_core.py @@ -14,7 +14,7 @@ from sglang.srt.mem_cache.base_prefix_cache import CacheRequestHandle from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class TestDecodeHiCacheTreeCore(CustomTestCase): diff --git a/test/registered/unit/disaggregation/test_decode_queue_cleanup.py b/test/registered/unit/disaggregation/test_decode_queue_cleanup.py index ff0c001fd1..f47c8ef7fd 100644 --- a/test/registered/unit/disaggregation/test_decode_queue_cleanup.py +++ b/test/registered/unit/disaggregation/test_decode_queue_cleanup.py @@ -21,7 +21,7 @@ from sglang.test.separate_buffer_allocator_double import ( ) from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class FakeReceiver: diff --git a/test/registered/unit/disaggregation/test_decode_req_to_token_pool.py b/test/registered/unit/disaggregation/test_decode_req_to_token_pool.py index 654afe253b..70c9e4c828 100644 --- a/test/registered/unit/disaggregation/test_decode_req_to_token_pool.py +++ b/test/registered/unit/disaggregation/test_decode_req_to_token_pool.py @@ -9,7 +9,7 @@ from sglang.srt.disaggregation.decode import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") def _init_decode_pool(pool): diff --git a/test/registered/unit/disaggregation/test_deferred_decode_kv_release.py b/test/registered/unit/disaggregation/test_deferred_decode_kv_release.py index 0a8936f219..b6c3f8c52e 100644 --- a/test/registered/unit/disaggregation/test_deferred_decode_kv_release.py +++ b/test/registered/unit/disaggregation/test_deferred_decode_kv_release.py @@ -18,7 +18,7 @@ from sglang.srt.disaggregation.decode import DecodeTransferQueue from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=17, suite="base-a-test-cpu") def _make_manager(): diff --git a/test/registered/unit/disaggregation/test_disaggregation_wire.py b/test/registered/unit/disaggregation/test_disaggregation_wire.py index 0a9bc35d4e..20f83cb47d 100644 --- a/test/registered/unit/disaggregation/test_disaggregation_wire.py +++ b/test/registered/unit/disaggregation/test_disaggregation_wire.py @@ -64,7 +64,7 @@ from sglang.srt.speculative.eagle_disaggregation import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=24, suite="base-a-test-cpu") class TestDisaggregationWire(unittest.TestCase): diff --git a/test/registered/unit/disaggregation/test_encode_receiver.py b/test/registered/unit/disaggregation/test_encode_receiver.py index 8e201e805f..a035304410 100644 --- a/test/registered/unit/disaggregation/test_encode_receiver.py +++ b/test/registered/unit/disaggregation/test_encode_receiver.py @@ -24,7 +24,7 @@ from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_registration_request(request_cls): diff --git a/test/registered/unit/disaggregation/test_encode_server.py b/test/registered/unit/disaggregation/test_encode_server.py index 6f0c093dbf..4282db9a6e 100644 --- a/test/registered/unit/disaggregation/test_encode_server.py +++ b/test/registered/unit/disaggregation/test_encode_server.py @@ -53,7 +53,7 @@ from sglang.srt.utils.common import safe_pickle_loads from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestEncoderDPErrorHandling(CustomTestCase): diff --git a/test/registered/unit/disaggregation/test_fake_kv_sender.py b/test/registered/unit/disaggregation/test_fake_kv_sender.py index f8949bb910..7fcaa1ac1c 100644 --- a/test/registered/unit/disaggregation/test_fake_kv_sender.py +++ b/test/registered/unit/disaggregation/test_fake_kv_sender.py @@ -10,7 +10,7 @@ from sglang.srt.disaggregation.utils import DisaggregationMode from sglang.srt.environ import envs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestFakeKVSender(unittest.TestCase): diff --git a/test/registered/unit/disaggregation/test_kimi_k3_encoder_mode.py b/test/registered/unit/disaggregation/test_kimi_k3_encoder_mode.py index beb3eaaa68..286935f23e 100644 --- a/test/registered/unit/disaggregation/test_kimi_k3_encoder_mode.py +++ b/test/registered/unit/disaggregation/test_kimi_k3_encoder_mode.py @@ -56,7 +56,7 @@ from sglang.srt.server_args import resolve_encoder_transfer_backend from sglang.srt.utils import ImageData from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def test_kimi_k3_encoder_transfer_backend_auto_avoids_tp_fanout(): diff --git a/test/registered/unit/disaggregation/test_kv_events.py b/test/registered/unit/disaggregation/test_kv_events.py index 077cda160e..098529fe94 100644 --- a/test/registered/unit/disaggregation/test_kv_events.py +++ b/test/registered/unit/disaggregation/test_kv_events.py @@ -24,7 +24,7 @@ from sglang.srt.disaggregation.kv_events import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestResolveLoadPubRange(CustomTestCase): diff --git a/test/registered/unit/disaggregation/test_kv_transfer_replica_metric.py b/test/registered/unit/disaggregation/test_kv_transfer_replica_metric.py index 75baa3f52a..78dd6ec7cd 100644 --- a/test/registered/unit/disaggregation/test_kv_transfer_replica_metric.py +++ b/test/registered/unit/disaggregation/test_kv_transfer_replica_metric.py @@ -17,7 +17,7 @@ from sglang.srt.disaggregation.common.conn import CommonKVManager, CommonKVSende from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") KV_ITEM_LENS_SUM = 100 STATE_ITEM_LENS_SUM = 7 diff --git a/test/registered/unit/disaggregation/test_minimax_sparse_disagg_state_kv_args.py b/test/registered/unit/disaggregation/test_minimax_sparse_disagg_state_kv_args.py index acb8c3830d..e0e8d808ae 100644 --- a/test/registered/unit/disaggregation/test_minimax_sparse_disagg_state_kv_args.py +++ b/test/registered/unit/disaggregation/test_minimax_sparse_disagg_state_kv_args.py @@ -7,7 +7,7 @@ from sglang.srt.disaggregation.utils import setup_state_kv_args from sglang.srt.mem_cache.memory_pool import MiniMaxSparseKVPool from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") def _make_k_only_pool(start_layer: int = 0) -> MiniMaxSparseKVPool: diff --git a/test/registered/unit/disaggregation/test_mooncake_transfer_batching.py b/test/registered/unit/disaggregation/test_mooncake_transfer_batching.py index d22af73993..5c719a9725 100644 --- a/test/registered/unit/disaggregation/test_mooncake_transfer_batching.py +++ b/test/registered/unit/disaggregation/test_mooncake_transfer_batching.py @@ -12,7 +12,7 @@ from sglang.srt.disaggregation.mooncake.conn import MooncakeKVManager from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestMooncakeTransferBatching(unittest.TestCase): diff --git a/test/registered/unit/disaggregation/test_nixl_backend_basic.py b/test/registered/unit/disaggregation/test_nixl_backend_basic.py index 2bf28ec9bd..771c8ad4d6 100644 --- a/test/registered/unit/disaggregation/test_nixl_backend_basic.py +++ b/test/registered/unit/disaggregation/test_nixl_backend_basic.py @@ -29,7 +29,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class NotificationFakeAgent: diff --git a/test/registered/unit/disaggregation/test_nixl_deferred_kv_release.py b/test/registered/unit/disaggregation/test_nixl_deferred_kv_release.py index 752787665e..d80e97549b 100644 --- a/test/registered/unit/disaggregation/test_nixl_deferred_kv_release.py +++ b/test/registered/unit/disaggregation/test_nixl_deferred_kv_release.py @@ -16,7 +16,7 @@ from sglang.srt.disaggregation.nixl.conn import NixlKVManager from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _prefill_mgr(cls=CommonKVManager, enabled=True): diff --git a/test/registered/unit/disaggregation/test_pp_hybrid_kv_transfer.py b/test/registered/unit/disaggregation/test_pp_hybrid_kv_transfer.py index 1fd5a9417d..ee36cfbb63 100644 --- a/test/registered/unit/disaggregation/test_pp_hybrid_kv_transfer.py +++ b/test/registered/unit/disaggregation/test_pp_hybrid_kv_transfer.py @@ -21,7 +21,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _full_attention_ids(*, num_layers: int, interval: int) -> list: diff --git a/test/registered/unit/disaggregation/test_pp_mla_kv_transfer.py b/test/registered/unit/disaggregation/test_pp_mla_kv_transfer.py index 0c6034705e..c790496ae1 100644 --- a/test/registered/unit/disaggregation/test_pp_mla_kv_transfer.py +++ b/test/registered/unit/disaggregation/test_pp_mla_kv_transfer.py @@ -9,7 +9,7 @@ from sglang.srt.disaggregation.nixl.conn import NixlKVManager from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") NUM_LAYERS = 78 ITEM_LEN = 576 diff --git a/test/registered/unit/disaggregation/test_prefill_abort_result_cleanup.py b/test/registered/unit/disaggregation/test_prefill_abort_result_cleanup.py index 4286a29f55..e4788bfb79 100644 --- a/test/registered/unit/disaggregation/test_prefill_abort_result_cleanup.py +++ b/test/registered/unit/disaggregation/test_prefill_abort_result_cleanup.py @@ -18,7 +18,7 @@ from sglang.srt.mem_cache.base_prefix_cache import ( from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _Req: diff --git a/test/registered/unit/disaggregation/test_receiver_connection_pool.py b/test/registered/unit/disaggregation/test_receiver_connection_pool.py index 9cdbcf8a20..a20cd369b5 100644 --- a/test/registered/unit/disaggregation/test_receiver_connection_pool.py +++ b/test/registered/unit/disaggregation/test_receiver_connection_pool.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") import threading diff --git a/test/registered/unit/disaggregation/test_register_to_bootstrap.py b/test/registered/unit/disaggregation/test_register_to_bootstrap.py index 312040e8d9..3b73025845 100644 --- a/test/registered/unit/disaggregation/test_register_to_bootstrap.py +++ b/test/registered/unit/disaggregation/test_register_to_bootstrap.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/disaggregation/test_specv2_kvcache_offloading.py b/test/registered/unit/disaggregation/test_specv2_kvcache_offloading.py index 89d094418c..28a9ea0d98 100644 --- a/test/registered/unit/disaggregation/test_specv2_kvcache_offloading.py +++ b/test/registered/unit/disaggregation/test_specv2_kvcache_offloading.py @@ -32,7 +32,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _make_mock_req( diff --git a/test/registered/unit/disaggregation/test_staging_draft_kv_slots.py b/test/registered/unit/disaggregation/test_staging_draft_kv_slots.py index 8350b5d54c..91e4d3c25d 100644 --- a/test/registered/unit/disaggregation/test_staging_draft_kv_slots.py +++ b/test/registered/unit/disaggregation/test_staging_draft_kv_slots.py @@ -17,7 +17,7 @@ from sglang.srt.mem_cache.memory_pool import HybridLinearKVPool, MHATokenToKVPoo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _Pool(MHATokenToKVPool): diff --git a/test/registered/unit/disaggregation/test_unified_memory_move_gate.py b/test/registered/unit/disaggregation/test_unified_memory_move_gate.py index e50cd4ff85..f295c45712 100644 --- a/test/registered/unit/disaggregation/test_unified_memory_move_gate.py +++ b/test/registered/unit/disaggregation/test_unified_memory_move_gate.py @@ -25,7 +25,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeTransferQueue: diff --git a/test/registered/unit/distributed/test_gated_launch.py b/test/registered/unit/distributed/test_gated_launch.py index 8e583cd309..eaf5f256f4 100644 --- a/test/registered/unit/distributed/test_gated_launch.py +++ b/test/registered/unit/distributed/test_gated_launch.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=34, suite="base-a-test-cpu") +register_cpu_ci(est_time=17, suite="base-a-test-cpu") import multiprocessing import threading diff --git a/test/registered/unit/distributed/test_get_default_distributed_backend.py b/test/registered/unit/distributed/test_get_default_distributed_backend.py index 8681c5dc5b..fa6bf9ba2b 100644 --- a/test/registered/unit/distributed/test_get_default_distributed_backend.py +++ b/test/registered/unit/distributed/test_get_default_distributed_backend.py @@ -8,7 +8,7 @@ from sglang.srt.platforms.interface import SRTPlatform from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") # Worked example: an out-of-tree platform that overrides the torch backend. diff --git a/test/registered/unit/distributed/test_parallel_state.py b/test/registered/unit/distributed/test_parallel_state.py index edb86d6547..b289eabcc0 100644 --- a/test/registered/unit/distributed/test_parallel_state.py +++ b/test/registered/unit/distributed/test_parallel_state.py @@ -46,7 +46,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import publish_build_topology -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # Import the actual parallel_state module parallel_state = pytest.importorskip("sglang.srt.distributed.parallel_state") diff --git a/test/registered/unit/distributed/test_pp_comm_overlap.py b/test/registered/unit/distributed/test_pp_comm_overlap.py index 2243ef754f..2be3ed0371 100644 --- a/test/registered/unit/distributed/test_pp_comm_overlap.py +++ b/test/registered/unit/distributed/test_pp_comm_overlap.py @@ -10,7 +10,7 @@ from sglang.srt.managers.scheduler_pp_mixin import SchedulerPPMixin from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class FakeStream: diff --git a/test/registered/unit/entrypoints/anthropic/test_serving.py b/test/registered/unit/entrypoints/anthropic/test_serving.py index 96a990bcc0..812a1f4044 100644 --- a/test/registered/unit/entrypoints/anthropic/test_serving.py +++ b/test/registered/unit/entrypoints/anthropic/test_serving.py @@ -24,7 +24,7 @@ from sglang.srt.parser.template_detection import ( # noqa: E402 ) from sglang.test.ci.ci_register import register_cpu_ci # noqa: E402 -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class _FakeOpenAIServingChat: diff --git a/test/registered/unit/entrypoints/openai/test_audio_chunking.py b/test/registered/unit/entrypoints/openai/test_audio_chunking.py index ff87e974ed..c850668147 100644 --- a/test/registered/unit/entrypoints/openai/test_audio_chunking.py +++ b/test/registered/unit/entrypoints/openai/test_audio_chunking.py @@ -24,7 +24,7 @@ from sglang.srt.entrypoints.openai.audio_chunking import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") SR = 16000 diff --git a/test/registered/unit/entrypoints/openai/test_chat_prompt_round_trip.py b/test/registered/unit/entrypoints/openai/test_chat_prompt_round_trip.py index bb398dd372..7d9b625082 100644 --- a/test/registered/unit/entrypoints/openai/test_chat_prompt_round_trip.py +++ b/test/registered/unit/entrypoints/openai/test_chat_prompt_round_trip.py @@ -12,7 +12,7 @@ from types import SimpleNamespace from sglang.srt.entrypoints.openai.serving_chat import OpenAIServingChat from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") probe = OpenAIServingChat._probe_prompt_text_round_trip engine_prompt = OpenAIServingChat._engine_prompt diff --git a/test/registered/unit/entrypoints/openai/test_exa_search.py b/test/registered/unit/entrypoints/openai/test_exa_search.py index 8b7cb06b91..3578e7e3a2 100644 --- a/test/registered/unit/entrypoints/openai/test_exa_search.py +++ b/test/registered/unit/entrypoints/openai/test_exa_search.py @@ -16,7 +16,7 @@ from sglang.srt.entrypoints.search.exa_client import ( from sglang.srt.entrypoints.tool import HarmonyBrowserTool from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class ExaClientTestCase(unittest.TestCase): diff --git a/test/registered/unit/entrypoints/openai/test_matched_stop.py b/test/registered/unit/entrypoints/openai/test_matched_stop.py index 9e3e950e05..b23b07173e 100644 --- a/test/registered/unit/entrypoints/openai/test_matched_stop.py +++ b/test/registered/unit/entrypoints/openai/test_matched_stop.py @@ -3,7 +3,7 @@ import unittest from sglang.srt.sampling.sampling_params import MAX_LEN, get_max_seq_length from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestRegexPatternMaxLength(unittest.TestCase): diff --git a/test/registered/unit/entrypoints/openai/test_protocol.py b/test/registered/unit/entrypoints/openai/test_protocol.py index c580e691f4..c40c11f90a 100644 --- a/test/registered/unit/entrypoints/openai/test_protocol.py +++ b/test/registered/unit/entrypoints/openai/test_protocol.py @@ -36,7 +36,7 @@ from sglang.srt.entrypoints.openai.protocol import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestModelCard(unittest.TestCase): diff --git a/test/registered/unit/entrypoints/openai/test_responses_custom_tools.py b/test/registered/unit/entrypoints/openai/test_responses_custom_tools.py index da8a3116de..f477545df7 100644 --- a/test/registered/unit/entrypoints/openai/test_responses_custom_tools.py +++ b/test/registered/unit/entrypoints/openai/test_responses_custom_tools.py @@ -24,7 +24,7 @@ from sglang.srt.entrypoints.openai.serving_responses import OpenAIServingRespons from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") CUSTOM_TOOL = { "type": "custom", diff --git a/test/registered/unit/entrypoints/openai/test_responses_protocol.py b/test/registered/unit/entrypoints/openai/test_responses_protocol.py index f186f31bb1..2941c0dfe1 100644 --- a/test/registered/unit/entrypoints/openai/test_responses_protocol.py +++ b/test/registered/unit/entrypoints/openai/test_responses_protocol.py @@ -12,7 +12,7 @@ from sglang.srt.entrypoints.openai.protocol import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") def _in_progress_response(request: ResponsesRequest) -> ResponsesResponse: diff --git a/test/registered/unit/entrypoints/openai/test_serving_chat.py b/test/registered/unit/entrypoints/openai/test_serving_chat.py index 343f163549..b347009683 100644 --- a/test/registered/unit/entrypoints/openai/test_serving_chat.py +++ b/test/registered/unit/entrypoints/openai/test_serving_chat.py @@ -53,7 +53,7 @@ from sglang.srt.server_args import ServerArgs from sglang.srt.utils import get_or_create_event_loop from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=112, suite="base-a-test-cpu") # Every spec resolve_chat_encoding_spec can return; pinned by the guard below. _ALL_CHAT_ENCODING_SPECS = ("dsv41", "dsv4", "dsv32", "inkling", "kimi_k3") diff --git a/test/registered/unit/entrypoints/openai/test_serving_completions.py b/test/registered/unit/entrypoints/openai/test_serving_completions.py index 2ce597b1fe..3ac2122058 100644 --- a/test/registered/unit/entrypoints/openai/test_serving_completions.py +++ b/test/registered/unit/entrypoints/openai/test_serving_completions.py @@ -24,7 +24,7 @@ from sglang.srt.server_args import ServerArgs from sglang.srt.utils import get_or_create_event_loop from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _spec_result(index): diff --git a/test/registered/unit/entrypoints/openai/test_serving_embedding.py b/test/registered/unit/entrypoints/openai/test_serving_embedding.py index 90d2f95877..c0db82fe61 100644 --- a/test/registered/unit/entrypoints/openai/test_serving_embedding.py +++ b/test/registered/unit/entrypoints/openai/test_serving_embedding.py @@ -58,7 +58,7 @@ from sglang.srt.entrypoints.openai.serving_embedding import OpenAIServingEmbeddi from sglang.srt.managers.io_struct import EmbeddingReqInput from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") register_cpu_ci(est_time=6, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/entrypoints/openai/test_serving_transcription.py b/test/registered/unit/entrypoints/openai/test_serving_transcription.py index 531a7a0780..90e2bae847 100644 --- a/test/registered/unit/entrypoints/openai/test_serving_transcription.py +++ b/test/registered/unit/entrypoints/openai/test_serving_transcription.py @@ -38,7 +38,7 @@ from sglang.srt.utils import get_or_create_event_loop from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _chunk(text: str, finish: str = None) -> dict: diff --git a/test/registered/unit/entrypoints/openai/test_transcription_adapters.py b/test/registered/unit/entrypoints/openai/test_transcription_adapters.py index 571f49b9de..0184064688 100644 --- a/test/registered/unit/entrypoints/openai/test_transcription_adapters.py +++ b/test/registered/unit/entrypoints/openai/test_transcription_adapters.py @@ -24,7 +24,7 @@ from sglang.srt.entrypoints.openai.transcription_adapters import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # Per-second scaling rate every speech-LM adapter uses for max_new_tokens. diff --git a/test/registered/unit/entrypoints/openai/test_whisper_adapter.py b/test/registered/unit/entrypoints/openai/test_whisper_adapter.py index 4abcedffa2..da60fc7a5f 100644 --- a/test/registered/unit/entrypoints/openai/test_whisper_adapter.py +++ b/test/registered/unit/entrypoints/openai/test_whisper_adapter.py @@ -21,7 +21,7 @@ from sglang.srt.entrypoints.openai.transcription_adapters.whisper import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") class TestWhisperParseFusedOutput(CustomTestCase): diff --git a/test/registered/unit/entrypoints/test_grpc_bridge.py b/test/registered/unit/entrypoints/test_grpc_bridge.py index b48abe8b9f..97baabbdd3 100644 --- a/test/registered/unit/entrypoints/test_grpc_bridge.py +++ b/test/registered/unit/entrypoints/test_grpc_bridge.py @@ -8,7 +8,7 @@ from sglang.srt.managers.tokenizer_manager import ServerStatus, TokenizerManager from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _ChunkStatus(enum.Enum): diff --git a/test/registered/unit/entrypoints/test_http_server_warmup.py b/test/registered/unit/entrypoints/test_http_server_warmup.py index bfc3a8b48f..0a4515f42e 100644 --- a/test/registered/unit/entrypoints/test_http_server_warmup.py +++ b/test/registered/unit/entrypoints/test_http_server_warmup.py @@ -9,7 +9,7 @@ from sglang.srt.entrypoints.http_server import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDisaggregationServerWarmup(unittest.IsolatedAsyncioTestCase): diff --git a/test/registered/unit/entrypoints/test_ready.py b/test/registered/unit/entrypoints/test_ready.py index ebe960d634..5455f26259 100644 --- a/test/registered/unit/entrypoints/test_ready.py +++ b/test/registered/unit/entrypoints/test_ready.py @@ -7,7 +7,7 @@ from sglang.srt.managers.tokenizer_manager import ServerStatus, TokenizerManager from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestReadyEndpoint(CustomTestCase): diff --git a/test/registered/unit/entrypoints/test_rust_server_dp_ports.py b/test/registered/unit/entrypoints/test_rust_server_dp_ports.py index a834d7b08b..8c5c1c3ea7 100644 --- a/test/registered/unit/entrypoints/test_rust_server_dp_ports.py +++ b/test/registered/unit/entrypoints/test_rust_server_dp_ports.py @@ -9,7 +9,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.srt.rust_server import server as rust_server from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") @pytest.mark.parametrize( diff --git a/test/registered/unit/entrypoints/test_server_info.py b/test/registered/unit/entrypoints/test_server_info.py index f159dec053..d8ae922a34 100644 --- a/test/registered/unit/entrypoints/test_server_info.py +++ b/test/registered/unit/entrypoints/test_server_info.py @@ -37,7 +37,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") class _CustomModelLoader: diff --git a/test/registered/unit/entrypoints/test_server_warmup.py b/test/registered/unit/entrypoints/test_server_warmup.py index 824b881985..e771ae82ff 100644 --- a/test/registered/unit/entrypoints/test_server_warmup.py +++ b/test/registered/unit/entrypoints/test_server_warmup.py @@ -13,7 +13,7 @@ from sglang.srt.entrypoints.http_server import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestVlmWarmupImage(CustomTestCase): diff --git a/test/registered/unit/entrypoints/test_ssl_cert_refresher.py b/test/registered/unit/entrypoints/test_ssl_cert_refresher.py index 6773848485..d526f2b7be 100644 --- a/test/registered/unit/entrypoints/test_ssl_cert_refresher.py +++ b/test/registered/unit/entrypoints/test_ssl_cert_refresher.py @@ -8,7 +8,7 @@ from sglang.srt.entrypoints.ssl_utils import SSLCertRefresher from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=17, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") def _make_temp_pem(content: bytes) -> str: diff --git a/test/registered/unit/entrypoints/test_v1_loads_aggregate.py b/test/registered/unit/entrypoints/test_v1_loads_aggregate.py index ad3ee9247e..432ecab038 100644 --- a/test/registered/unit/entrypoints/test_v1_loads_aggregate.py +++ b/test/registered/unit/entrypoints/test_v1_loads_aggregate.py @@ -31,7 +31,7 @@ from sglang.test.test_utils import CustomTestCase, maybe_stub_sgl_kernel maybe_stub_sgl_kernel() -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _temp_path() -> str: diff --git a/test/registered/unit/eplb/test_balanced_packing.py b/test/registered/unit/eplb/test_balanced_packing.py index 2304dc82c8..d5950333b6 100644 --- a/test/registered/unit/eplb/test_balanced_packing.py +++ b/test/registered/unit/eplb/test_balanced_packing.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py b/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py index 973268ff36..e6a83fb88f 100644 --- a/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py +++ b/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/eplb/test_dispatch_dtype_preservation.py b/test/registered/unit/eplb/test_dispatch_dtype_preservation.py index 80c822a4df..f09bbca961 100644 --- a/test/registered/unit/eplb/test_dispatch_dtype_preservation.py +++ b/test/registered/unit/eplb/test_dispatch_dtype_preservation.py @@ -9,7 +9,7 @@ a specific dtype (int32). from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/eplb/test_waterfill_eplb.py b/test/registered/unit/eplb/test_waterfill_eplb.py index 99503d43e2..05b388ee0e 100644 --- a/test/registered/unit/eplb/test_waterfill_eplb.py +++ b/test/registered/unit/eplb/test_waterfill_eplb.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/function_call/test_deepseekv41_detector.py b/test/registered/unit/function_call/test_deepseekv41_detector.py index a7a960d953..74ec9aa370 100644 --- a/test/registered/unit/function_call/test_deepseekv41_detector.py +++ b/test/registered/unit/function_call/test_deepseekv41_detector.py @@ -21,7 +21,7 @@ from sglang.srt.parser.reasoning_parser import ReasoningParser from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") CHUNK_SIZES = [1, 2, 3, 5, 7, 11, 23, 1000] DSML = "|DSML|" diff --git a/test/registered/unit/function_call/test_dots_detector.py b/test/registered/unit/function_call/test_dots_detector.py index ef40cd1774..8f6845a44c 100644 --- a/test/registered/unit/function_call/test_dots_detector.py +++ b/test/registered/unit/function_call/test_dots_detector.py @@ -7,7 +7,7 @@ from sglang.srt.function_call.function_call_parser import FunctionCallParser from sglang.srt.parser.reasoning_parser import Qwen3Detector, ReasoningParser from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") def _tool(name: str, properties: dict) -> Tool: diff --git a/test/registered/unit/function_call/test_function_call_parser.py b/test/registered/unit/function_call/test_function_call_parser.py index 359ae726db..cab399dbb5 100644 --- a/test/registered/unit/function_call/test_function_call_parser.py +++ b/test/registered/unit/function_call/test_function_call_parser.py @@ -40,7 +40,7 @@ from sglang.srt.function_call.qwen3_coder_detector import Qwen3CoderDetector from sglang.srt.function_call.utils import get_schema_properties from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") register_cpu_ci(est_time=70, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/function_call/test_glm47_schema_types.py b/test/registered/unit/function_call/test_glm47_schema_types.py index 8e37fc28da..5a8879924e 100644 --- a/test/registered/unit/function_call/test_glm47_schema_types.py +++ b/test/registered/unit/function_call/test_glm47_schema_types.py @@ -5,7 +5,7 @@ from sglang.srt.entrypoints.openai.protocol import Function, Tool from sglang.srt.function_call.glm47_moe_detector import Glm47MoeDetector from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestGlm47SchemaTypes(unittest.TestCase): diff --git a/test/registered/unit/function_call/test_hunyuan_detector.py b/test/registered/unit/function_call/test_hunyuan_detector.py index adba60d4b2..95b37618a9 100644 --- a/test/registered/unit/function_call/test_hunyuan_detector.py +++ b/test/registered/unit/function_call/test_hunyuan_detector.py @@ -8,7 +8,7 @@ from sglang.srt.function_call.hunyuan_detector import HunyuanDetector from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") def _make_tools(): diff --git a/test/registered/unit/function_call/test_k2_v3_detector.py b/test/registered/unit/function_call/test_k2_v3_detector.py index 85d9f61cb7..e0fdaccd6f 100644 --- a/test/registered/unit/function_call/test_k2_v3_detector.py +++ b/test/registered/unit/function_call/test_k2_v3_detector.py @@ -8,7 +8,7 @@ from sglang.srt.function_call.k2_v3_detector import K2V3Detector from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def make_weather_tool() -> Tool: diff --git a/test/registered/unit/function_call/test_kimik3_structural_tag.py b/test/registered/unit/function_call/test_kimik3_structural_tag.py index 9bef313ac8..f01ef2fcea 100644 --- a/test/registered/unit/function_call/test_kimik3_structural_tag.py +++ b/test/registered/unit/function_call/test_kimik3_structural_tag.py @@ -28,7 +28,7 @@ from sglang.srt.function_call.kimik3_structural_tag import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") _CLOSE_TOKEN = "<|close|>" _CLOSE_TOKEN_ID = 256 diff --git a/test/registered/unit/function_call/test_minimax_m3_detector.py b/test/registered/unit/function_call/test_minimax_m3_detector.py index 0530df560c..8f6179ea0b 100644 --- a/test/registered/unit/function_call/test_minimax_m3_detector.py +++ b/test/registered/unit/function_call/test_minimax_m3_detector.py @@ -6,7 +6,7 @@ from sglang.srt.function_call.minimax_m3 import MINIMAX_NS_TOKEN, MinimaxM3Detec from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") NS = MINIMAX_NS_TOKEN diff --git a/test/registered/unit/function_call/test_muse_glimmer_format.py b/test/registered/unit/function_call/test_muse_glimmer_format.py index 682bbb7d07..47a54121a6 100644 --- a/test/registered/unit/function_call/test_muse_glimmer_format.py +++ b/test/registered/unit/function_call/test_muse_glimmer_format.py @@ -24,7 +24,7 @@ from sglang.srt.function_call.muse_glimmer_format import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _CHANNEL_MARKERS = (MESSAGE, EOM, EOT, START) _HELD_BACK_MARKERS = (MESSAGE, EOM, EOT, START, FUNCTION_CALLS_OPEN, INVOKE_OPEN) diff --git a/test/registered/unit/function_call/test_spark25_detector.py b/test/registered/unit/function_call/test_spark25_detector.py index 32b75da266..a2858fba15 100644 --- a/test/registered/unit/function_call/test_spark25_detector.py +++ b/test/registered/unit/function_call/test_spark25_detector.py @@ -10,7 +10,7 @@ from sglang.srt.function_call.spark25_detector import Spark25Detector from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _xml(name: str, arguments: list[tuple[str, str]]) -> str: diff --git a/test/registered/unit/hardware_backend/mlx/test_mlx_quantization_override.py b/test/registered/unit/hardware_backend/mlx/test_mlx_quantization_override.py index 473464a75f..1df51d0562 100644 --- a/test/registered/unit/hardware_backend/mlx/test_mlx_quantization_override.py +++ b/test/registered/unit/hardware_backend/mlx/test_mlx_quantization_override.py @@ -11,7 +11,7 @@ import unittest from sglang.srt.layers.quantization.mlx import MlxQuantizationConfig from sglang.test.ci.ci_register import register_cpu_ci, register_mlx_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") register_mlx_ci(est_time=1, suite="stage-a-unit-test-mlx") diff --git a/test/registered/unit/hardware_backend/mlx/test_mlx_remote_code_gate.py b/test/registered/unit/hardware_backend/mlx/test_mlx_remote_code_gate.py index ff65e651fb..78175146c0 100644 --- a/test/registered/unit/hardware_backend/mlx/test_mlx_remote_code_gate.py +++ b/test/registered/unit/hardware_backend/mlx/test_mlx_remote_code_gate.py @@ -19,7 +19,7 @@ from unittest.mock import patch from sglang.test.ci.ci_register import register_cpu_ci, register_mlx_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_mlx_ci(est_time=5, suite="stage-a-unit-test-mlx") _HAS_MLX = ( diff --git a/test/registered/unit/kernels/test_fused_op_dispatch.py b/test/registered/unit/kernels/test_fused_op_dispatch.py index fd18fc951c..2bb8b20fd2 100644 --- a/test/registered/unit/kernels/test_fused_op_dispatch.py +++ b/test/registered/unit/kernels/test_fused_op_dispatch.py @@ -28,7 +28,7 @@ from sglang.kernels.spec import CapabilityRequirement as Cap from sglang.kernels.spec import KernelBackend, PlatformInfo from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=18, suite="base-a-test-cpu") _CUDA = PlatformInfo(device_type="cuda", cuda_arch_major=9, cuda_arch_minor=0) _HIP = PlatformInfo(device_type="hip") diff --git a/test/registered/unit/kernels/test_jit_cache.py b/test/registered/unit/kernels/test_jit_cache.py index d70ca03864..608e6c49e0 100644 --- a/test/registered/unit/kernels/test_jit_cache.py +++ b/test/registered/unit/kernels/test_jit_cache.py @@ -19,7 +19,7 @@ from sglang.kernels.jit.utils.compile.paths import KERNEL_PATH from sglang.kernels.jit.utils.compile.spec import BuildSpec from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") @pytest.fixture(autouse=True) diff --git a/test/registered/unit/layers/attention/linear/kernels/test_kda_nvidia.py b/test/registered/unit/layers/attention/linear/kernels/test_kda_nvidia.py index ffa952a586..2bdc6ea78d 100644 --- a/test/registered/unit/layers/attention/linear/kernels/test_kda_nvidia.py +++ b/test/registered/unit/layers/attention/linear/kernels/test_kda_nvidia.py @@ -13,7 +13,7 @@ from sglang.srt.layers.attention.linear.kernels.kda_nvidia import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _RejectTriton: diff --git a/test/registered/unit/layers/attention/linear/kernels/test_kda_ptx.py b/test/registered/unit/layers/attention/linear/kernels/test_kda_ptx.py index 5509c7ef70..4420a817bc 100644 --- a/test/registered/unit/layers/attention/linear/kernels/test_kda_ptx.py +++ b/test/registered/unit/layers/attention/linear/kernels/test_kda_ptx.py @@ -9,7 +9,7 @@ from sglang.srt.layers.attention.linear.kernels.kda_ptx import PtxKDAKernel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _RejectTriton: diff --git a/test/registered/unit/layers/attention/test_dots_hybrid_backend.py b/test/registered/unit/layers/attention/test_dots_hybrid_backend.py index 45ed6fe8f2..af9a31cfcd 100644 --- a/test/registered/unit/layers/attention/test_dots_hybrid_backend.py +++ b/test/registered/unit/layers/attention/test_dots_hybrid_backend.py @@ -22,7 +22,7 @@ from sglang.srt.layers.attention.swa_mla_fallback.ops import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _batch(*, bs: int, num_tokens: int, original_bs: int | None = None): diff --git a/test/registered/unit/layers/attention/test_dsa_head_gate_guard.py b/test/registered/unit/layers/attention/test_dsa_head_gate_guard.py index 246b24c4a6..9489a5b452 100644 --- a/test/registered/unit/layers/attention/test_dsa_head_gate_guard.py +++ b/test/registered/unit/layers/attention/test_dsa_head_gate_guard.py @@ -4,7 +4,7 @@ from sglang.srt.utils import is_cuda, is_hip from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") HELPERS = ("logits_head_gate_graph", "scale_head_gate_graph") diff --git a/test/registered/unit/layers/attention/test_dsa_mqa_logits_chunking.py b/test/registered/unit/layers/attention/test_dsa_mqa_logits_chunking.py index 4d9e0a703d..e0e7c8003b 100644 --- a/test/registered/unit/layers/attention/test_dsa_mqa_logits_chunking.py +++ b/test/registered/unit/layers/attention/test_dsa_mqa_logits_chunking.py @@ -19,7 +19,7 @@ from sglang.srt.layers.attention.mqa_logits_utils import ( # noqa: E402 ) from sglang.test.ci.ci_register import register_cpu_ci # noqa: E402 -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") CEILING = MQA_LOGITS_MAX_BYTES_ROCM # More than any single logits tensor here needs, so it never decides a case. diff --git a/test/registered/unit/layers/attention/test_encoder_decoder_varlen_gather.py b/test/registered/unit/layers/attention/test_encoder_decoder_varlen_gather.py index 1c0647d340..b89cd95fd1 100644 --- a/test/registered/unit/layers/attention/test_encoder_decoder_varlen_gather.py +++ b/test/registered/unit/layers/attention/test_encoder_decoder_varlen_gather.py @@ -22,7 +22,7 @@ with patch.dict( from sglang.srt.layers.attention import xpu_backend from sglang.srt.layers.attention.xpu_backend import XPUAttentionBackend -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestEncoderDecoderForward(unittest.TestCase): diff --git a/test/registered/unit/layers/attention/test_flashattention_graph_metadata.py b/test/registered/unit/layers/attention/test_flashattention_graph_metadata.py index ea750ff195..6ca7cc3308 100644 --- a/test/registered/unit/layers/attention/test_flashattention_graph_metadata.py +++ b/test/registered/unit/layers/attention/test_flashattention_graph_metadata.py @@ -9,7 +9,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") class TestFlashAttentionGraphMetadata(CustomTestCase): diff --git a/test/registered/unit/layers/attention/test_flashattention_pa_swa_prefill_lens_size.py b/test/registered/unit/layers/attention/test_flashattention_pa_swa_prefill_lens_size.py index 43e66379db..b7e35c4fe2 100644 --- a/test/registered/unit/layers/attention/test_flashattention_pa_swa_prefill_lens_size.py +++ b/test/registered/unit/layers/attention/test_flashattention_pa_swa_prefill_lens_size.py @@ -30,7 +30,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") class _FakeShardPool(PageInterleaveKVPoolMixin): diff --git a/test/registered/unit/layers/attention/test_gdn_flashinfer_alignment.py b/test/registered/unit/layers/attention/test_gdn_flashinfer_alignment.py index 1e55579fe6..02671e1306 100644 --- a/test/registered/unit/layers/attention/test_gdn_flashinfer_alignment.py +++ b/test/registered/unit/layers/attention/test_gdn_flashinfer_alignment.py @@ -8,7 +8,7 @@ from sglang.srt.layers.attention.linear.kernels.gdn_flashinfer import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") def _view_with_pointer_mod( diff --git a/test/registered/unit/layers/attention/test_gdn_prefill_backend_policy.py b/test/registered/unit/layers/attention/test_gdn_prefill_backend_policy.py index 182ee17a86..2af3cb8a38 100644 --- a/test/registered/unit/layers/attention/test_gdn_prefill_backend_policy.py +++ b/test/registered/unit/layers/attention/test_gdn_prefill_backend_policy.py @@ -25,7 +25,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _publish(testcase, **fields): diff --git a/test/registered/unit/layers/attention/test_index_topk_share.py b/test/registered/unit/layers/attention/test_index_topk_share.py index f92a2bb340..4735b1d74c 100644 --- a/test/registered/unit/layers/attention/test_index_topk_share.py +++ b/test/registered/unit/layers/attention/test_index_topk_share.py @@ -8,7 +8,7 @@ from sglang.srt.layers.attention.index_topk_share import IndexTopKShareState from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _batch( diff --git a/test/registered/unit/layers/attention/test_kda_helion_dispatcher.py b/test/registered/unit/layers/attention/test_kda_helion_dispatcher.py index e9f6464234..23176620b3 100644 --- a/test/registered/unit/layers/attention/test_kda_helion_dispatcher.py +++ b/test/registered/unit/layers/attention/test_kda_helion_dispatcher.py @@ -12,7 +12,7 @@ from sglang.srt.runtime_context import override_platform from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHelionKDADispatcher(unittest.TestCase): diff --git a/test/registered/unit/layers/attention/test_kv_shard_hooks.py b/test/registered/unit/layers/attention/test_kv_shard_hooks.py index f1ba777752..cf8d91e80a 100644 --- a/test/registered/unit/layers/attention/test_kv_shard_hooks.py +++ b/test/registered/unit/layers/attention/test_kv_shard_hooks.py @@ -26,7 +26,7 @@ from sglang.srt.mem_cache.page_interleave_pool import PageInterleaveKVPoolMixin from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _RecordingPool(PageInterleaveKVPoolMixin): diff --git a/test/registered/unit/layers/attention/test_kv_translate_ownership.py b/test/registered/unit/layers/attention/test_kv_translate_ownership.py index 45ef311497..11535b3dd8 100644 --- a/test/registered/unit/layers/attention/test_kv_translate_ownership.py +++ b/test/registered/unit/layers/attention/test_kv_translate_ownership.py @@ -37,7 +37,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") # The attention package is a namespace package (no __init__), so anchor the # scan on a concrete module inside it. diff --git a/test/registered/unit/layers/attention/test_linear_attn_config.py b/test/registered/unit/layers/attention/test_linear_attn_config.py index 57eb734de9..0f255a266e 100644 --- a/test/registered/unit/layers/attention/test_linear_attn_config.py +++ b/test/registered/unit/layers/attention/test_linear_attn_config.py @@ -25,7 +25,7 @@ from sglang.srt.layers.attention.linear.utils import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _Runner: diff --git a/test/registered/unit/layers/attention/test_mamba_track_state_dtype.py b/test/registered/unit/layers/attention/test_mamba_track_state_dtype.py index 879fe1df1b..c70b467036 100644 --- a/test/registered/unit/layers/attention/test_mamba_track_state_dtype.py +++ b/test/registered/unit/layers/attention/test_mamba_track_state_dtype.py @@ -9,7 +9,7 @@ from sglang.srt.layers.attention.mamba.mamba2_metadata import ForwardMetadata from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # Above the fp16 midpoint 1 + 2^-11 (single-rounds up to 1 + 2^-10) but below # the bf16 midpoint 1 + 2^-8 (rounds to 1.0, which then stays 1.0 in fp16): diff --git a/test/registered/unit/layers/attention/test_mla_decode_geometry.py b/test/registered/unit/layers/attention/test_mla_decode_geometry.py index b76f95d6bd..afce311952 100644 --- a/test/registered/unit/layers/attention/test_mla_decode_geometry.py +++ b/test/registered/unit/layers/attention/test_mla_decode_geometry.py @@ -45,7 +45,7 @@ from sglang.srt.environ import envs from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") # gfx950 full-GPU. Passed in rather than read from the device so this stays a CPU test. CORE_COUNT = 256 diff --git a/test/registered/unit/layers/attention/test_verify_mask.py b/test/registered/unit/layers/attention/test_verify_mask.py index ac8a21b0f0..d2eb4ec39b 100644 --- a/test/registered/unit/layers/attention/test_verify_mask.py +++ b/test/registered/unit/layers/attention/test_verify_mask.py @@ -15,7 +15,7 @@ from sglang.srt.speculative.eagle_utils import TreeMaskMode, default_tree_mask_m from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _MAX_BS = 4 _DRAFT = 3 diff --git a/test/registered/unit/layers/attention/test_vision_backend_selection.py b/test/registered/unit/layers/attention/test_vision_backend_selection.py index 3a91eaed88..e5261086af 100644 --- a/test/registered/unit/layers/attention/test_vision_backend_selection.py +++ b/test/registered/unit/layers/attention/test_vision_backend_selection.py @@ -10,7 +10,7 @@ from einops import rearrange from sglang.srt.layers.attention import vision from sglang.test.ci.ci_register import register_cpu_ci, register_npu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") register_npu_ci(est_time=2, suite="base-b-test-1-npu-a3") register_npu_ci(est_time=2, suite="nightly-1-npu-a3", nightly=True) diff --git a/test/registered/unit/layers/attention/test_vision_max_seqlen.py b/test/registered/unit/layers/attention/test_vision_max_seqlen.py index 4b9449182b..4794859893 100644 --- a/test/registered/unit/layers/attention/test_vision_max_seqlen.py +++ b/test/registered/unit/layers/attention/test_vision_max_seqlen.py @@ -9,7 +9,7 @@ from sglang.srt.models import kimi_k25 from sglang.srt.models.kimi_k25 import MoonViT3dEncoder, MoonViTEncoderLayer from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def test_vision_flash3_uses_precomputed_max_seqlen(monkeypatch): diff --git a/test/registered/unit/layers/attention/test_vision_strided_qkv.py b/test/registered/unit/layers/attention/test_vision_strided_qkv.py index 666d8d6a6c..e1c58ac9cf 100644 --- a/test/registered/unit/layers/attention/test_vision_strided_qkv.py +++ b/test/registered/unit/layers/attention/test_vision_strided_qkv.py @@ -18,7 +18,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import publish_build_topology -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") EMBED_DIM = 32 NUM_HEADS = 4 diff --git a/test/registered/unit/layers/moe/test_deepep_v2_buffer_lifecycle.py b/test/registered/unit/layers/moe/test_deepep_v2_buffer_lifecycle.py index b1ae19e530..766fd2f75f 100644 --- a/test/registered/unit/layers/moe/test_deepep_v2_buffer_lifecycle.py +++ b/test/registered/unit/layers/moe/test_deepep_v2_buffer_lifecycle.py @@ -10,7 +10,7 @@ from sglang.srt.runtime_context import get_resources, reset_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeGroup: diff --git a/test/registered/unit/layers/moe/test_deepep_v2_masked_slab.py b/test/registered/unit/layers/moe/test_deepep_v2_masked_slab.py index 61ed8ac97a..51db00c260 100644 --- a/test/registered/unit/layers/moe/test_deepep_v2_masked_slab.py +++ b/test/registered/unit/layers/moe/test_deepep_v2_masked_slab.py @@ -12,7 +12,7 @@ from sglang.kernels.ops.moe.ep_moe_kernels import ( from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-large") DEVICE = "cuda" diff --git a/test/registered/unit/layers/moe/test_deepep_v2_wire_dtype.py b/test/registered/unit/layers/moe/test_deepep_v2_wire_dtype.py index 47a0636b28..1bc8b6bbed 100644 --- a/test/registered/unit/layers/moe/test_deepep_v2_wire_dtype.py +++ b/test/registered/unit/layers/moe/test_deepep_v2_wire_dtype.py @@ -16,7 +16,7 @@ from sglang.srt.runtime_context import get_context, get_exec, reset_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=4, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") HIDDEN = 256 TOPK = 4 diff --git a/test/registered/unit/layers/moe/test_flashinfer_dispatcher.py b/test/registered/unit/layers/moe/test_flashinfer_dispatcher.py index e45b00b326..5b90b4e780 100644 --- a/test/registered/unit/layers/moe/test_flashinfer_dispatcher.py +++ b/test/registered/unit/layers/moe/test_flashinfer_dispatcher.py @@ -8,7 +8,7 @@ from sglang.srt.layers.moe.topk import StandardTopKOutput from sglang.srt.layers.moe.utils import FlashinferA2ADispatchType from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def test_empty_mxfp8_dispatch_uses_same_payload_dtype_as_nonempty_rank(): diff --git a/test/registered/unit/layers/moe/test_flashinfer_megamoe.py b/test/registered/unit/layers/moe/test_flashinfer_megamoe.py index cd73104ed9..bf3d42f51f 100644 --- a/test/registered/unit/layers/moe/test_flashinfer_megamoe.py +++ b/test/registered/unit/layers/moe/test_flashinfer_megamoe.py @@ -7,7 +7,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _load_megamoe_module(monkeypatch): diff --git a/test/registered/unit/layers/moe/test_fused_moe_native.py b/test/registered/unit/layers/moe/test_fused_moe_native.py index 90d74c235f..80e935979a 100644 --- a/test/registered/unit/layers/moe/test_fused_moe_native.py +++ b/test/registered/unit/layers/moe/test_fused_moe_native.py @@ -7,7 +7,7 @@ from sglang.srt.layers.moe.fused_moe_native import moe_forward_native from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestFusedMoeNative(CustomTestCase): diff --git a/test/registered/unit/layers/moe/test_fused_shared_expert_scaling.py b/test/registered/unit/layers/moe/test_fused_shared_expert_scaling.py index 4f3945e525..a1ba9efdf3 100644 --- a/test/registered/unit/layers/moe/test_fused_shared_expert_scaling.py +++ b/test/registered/unit/layers/moe/test_fused_shared_expert_scaling.py @@ -15,7 +15,7 @@ These tests pin the fused shared expert's topk weight contract on three paths: from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=61, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/layers/moe/test_mega_moe_deepgemm_api.py b/test/registered/unit/layers/moe/test_mega_moe_deepgemm_api.py index 445d567acc..c8ad3f7db1 100644 --- a/test/registered/unit/layers/moe/test_mega_moe_deepgemm_api.py +++ b/test/registered/unit/layers/moe/test_mega_moe_deepgemm_api.py @@ -16,7 +16,7 @@ maybe_stub_sgl_kernel() from sglang.srt.layers.moe import mega_moe from sglang.srt.runtime_context import get_parallel -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDeepGemmMegaMoeApi(CustomTestCase): diff --git a/test/registered/unit/layers/moe/test_topk_correction_bias_cache.py b/test/registered/unit/layers/moe/test_topk_correction_bias_cache.py index e1d282cbc6..2bf1bf10ef 100644 --- a/test/registered/unit/layers/moe/test_topk_correction_bias_cache.py +++ b/test/registered/unit/layers/moe/test_topk_correction_bias_cache.py @@ -10,7 +10,7 @@ from sglang.test.ci.ci_register import ( ) from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cuda_ci(est_time=10, stage="base-a", runner_config="1-gpu-small") register_amd_ci(est_time=5, stage="stage-a", runner_config="1-gpu-small-amd") diff --git a/test/registered/unit/layers/moe/test_w4afp8_deepep_dtype.py b/test/registered/unit/layers/moe/test_w4afp8_deepep_dtype.py index 4d28beb4c8..eee6d877ab 100644 --- a/test/registered/unit/layers/moe/test_w4afp8_deepep_dtype.py +++ b/test/registered/unit/layers/moe/test_w4afp8_deepep_dtype.py @@ -17,7 +17,7 @@ from sglang.srt.layers.quantization import w4afp8 from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestW4AFP8DeepEPDispatcherDtype(CustomTestCase): diff --git a/test/registered/unit/layers/moe/test_w4afp8_deepep_post_reorder.py b/test/registered/unit/layers/moe/test_w4afp8_deepep_post_reorder.py index 1a5414c91f..e2200d6658 100644 --- a/test/registered/unit/layers/moe/test_w4afp8_deepep_post_reorder.py +++ b/test/registered/unit/layers/moe/test_w4afp8_deepep_post_reorder.py @@ -10,7 +10,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # The function under test is a GPU implementation, but this test replaces every # launched kernel and only verifies the host-side call contract. Stub the diff --git a/test/registered/unit/layers/quantization/test_compressed_tensors_kv_cache.py b/test/registered/unit/layers/quantization/test_compressed_tensors_kv_cache.py index 54c6dbb895..39eac37454 100644 --- a/test/registered/unit/layers/quantization/test_compressed_tensors_kv_cache.py +++ b/test/registered/unit/layers/quantization/test_compressed_tensors_kv_cache.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/layers/quantization/test_compressed_tensors_lm_head.py b/test/registered/unit/layers/quantization/test_compressed_tensors_lm_head.py index 78bd0dc6f5..14198f4984 100644 --- a/test/registered/unit/layers/quantization/test_compressed_tensors_lm_head.py +++ b/test/registered/unit/layers/quantization/test_compressed_tensors_lm_head.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from unittest.mock import patch diff --git a/test/registered/unit/layers/quantization/test_compressed_tensors_mixed_precision.py b/test/registered/unit/layers/quantization/test_compressed_tensors_mixed_precision.py index 8cc3d347a2..9e4470c6b6 100644 --- a/test/registered/unit/layers/quantization/test_compressed_tensors_mixed_precision.py +++ b/test/registered/unit/layers/quantization/test_compressed_tensors_mixed_precision.py @@ -9,7 +9,7 @@ for every group. Both are config-parsing paths, so these tests run on CPU. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from unittest import mock diff --git a/test/registered/unit/layers/quantization/test_compressed_tensors_wna16_moe_no_linear.py b/test/registered/unit/layers/quantization/test_compressed_tensors_wna16_moe_no_linear.py index ebd356fdf0..f794230b89 100644 --- a/test/registered/unit/layers/quantization/test_compressed_tensors_wna16_moe_no_linear.py +++ b/test/registered/unit/layers/quantization/test_compressed_tensors_wna16_moe_no_linear.py @@ -16,7 +16,7 @@ from sglang.srt.runtime_context import override_platform from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _WNA16_MOE_SCHEMES = (CompressedTensorsWNA16MoE, CompressedTensorsWNA16TritonMoE) EXPERTS_LAYER = "model.layers.0.mlp.experts" diff --git a/test/registered/unit/layers/quantization/test_deepgemm_ue8m0_requant.py b/test/registered/unit/layers/quantization/test_deepgemm_ue8m0_requant.py index ddc039fcd4..b105b0605e 100644 --- a/test/registered/unit/layers/quantization/test_deepgemm_ue8m0_requant.py +++ b/test/registered/unit/layers/quantization/test_deepgemm_ue8m0_requant.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from unittest.mock import call, patch diff --git a/test/registered/unit/layers/quantization/test_flashinfer_trtllm_fp8_fallback.py b/test/registered/unit/layers/quantization/test_flashinfer_trtllm_fp8_fallback.py index 2ad08d33f9..f649abcffb 100644 --- a/test/registered/unit/layers/quantization/test_flashinfer_trtllm_fp8_fallback.py +++ b/test/registered/unit/layers/quantization/test_flashinfer_trtllm_fp8_fallback.py @@ -20,7 +20,7 @@ backend selector and the two GEMM implementations so they run on CPU CI. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import functools import unittest diff --git a/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py b/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py index a70bf545f3..26b844b928 100644 --- a/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py +++ b/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=35, suite="base-a-test-cpu") +register_cpu_ci(est_time=25, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/layers/quantization/test_fp8_blockwise_linear_backends.py b/test/registered/unit/layers/quantization/test_fp8_blockwise_linear_backends.py index 1d7dc98500..b3d3caa5f7 100644 --- a/test/registered/unit/layers/quantization/test_fp8_blockwise_linear_backends.py +++ b/test/registered/unit/layers/quantization/test_fp8_blockwise_linear_backends.py @@ -27,9 +27,9 @@ from sglang.test.layer_ut_utils import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=15, stage="base-b", runner_config="4-gpu-b200") -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") -register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=32, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=26, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=30, stage="base-b", runner_config="1-gpu-large") FP8_MAX = 448.0 diff --git a/test/registered/unit/layers/quantization/test_fp8_moe_runner_fallback.py b/test/registered/unit/layers/quantization/test_fp8_moe_runner_fallback.py index 4b3927e32c..df897ebe79 100644 --- a/test/registered/unit/layers/quantization/test_fp8_moe_runner_fallback.py +++ b/test/registered/unit/layers/quantization/test_fp8_moe_runner_fallback.py @@ -3,7 +3,7 @@ flashinfer_cutlass or flashinfer_cutedsl, which have no fp8 MoE path.""" from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from unittest.mock import patch diff --git a/test/registered/unit/layers/quantization/test_fp8_moe_runner_ownership.py b/test/registered/unit/layers/quantization/test_fp8_moe_runner_ownership.py index 55f6799410..6339c7dad9 100644 --- a/test/registered/unit/layers/quantization/test_fp8_moe_runner_ownership.py +++ b/test/registered/unit/layers/quantization/test_fp8_moe_runner_ownership.py @@ -24,7 +24,7 @@ from sglang.srt.runtime_context import get_flags from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _ACTIVATION_PARAMS = ("gemm1_alpha", "gemm1_beta", "gemm1_clamp_limit") diff --git a/test/registered/unit/layers/quantization/test_fp8_moe_weight_gating.py b/test/registered/unit/layers/quantization/test_fp8_moe_weight_gating.py index 85f4bb1ced..88f04ef463 100644 --- a/test/registered/unit/layers/quantization/test_fp8_moe_weight_gating.py +++ b/test/registered/unit/layers/quantization/test_fp8_moe_weight_gating.py @@ -9,7 +9,7 @@ follow the same shard count, or the two describe different tensors. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from unittest.mock import MagicMock, patch diff --git a/test/registered/unit/layers/quantization/test_humming_w4afp8_schemas.py b/test/registered/unit/layers/quantization/test_humming_w4afp8_schemas.py index 40cdc97476..ceeb076295 100644 --- a/test/registered/unit/layers/quantization/test_humming_w4afp8_schemas.py +++ b/test/registered/unit/layers/quantization/test_humming_w4afp8_schemas.py @@ -19,7 +19,7 @@ from sglang.srt.layers.quantization.humming import ( # noqa: E402 _W4AFp8CheckpointWeightSchema, ) -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestW4AFp8CheckpointSchema(CustomTestCase): diff --git a/test/registered/unit/layers/quantization/test_int8_linear_methods.py b/test/registered/unit/layers/quantization/test_int8_linear_methods.py index fa6f04329f..45f8322804 100644 --- a/test/registered/unit/layers/quantization/test_int8_linear_methods.py +++ b/test/registered/unit/layers/quantization/test_int8_linear_methods.py @@ -23,7 +23,7 @@ from sglang.test.layer_ut_utils import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-large") INT8_MAX = 127.0 diff --git a/test/registered/unit/layers/quantization/test_modelopt_nvfp4.py b/test/registered/unit/layers/quantization/test_modelopt_nvfp4.py index 8aeed7ae59..60602f8821 100644 --- a/test/registered/unit/layers/quantization/test_modelopt_nvfp4.py +++ b/test/registered/unit/layers/quantization/test_modelopt_nvfp4.py @@ -14,7 +14,7 @@ from sglang.srt.model_loader.weight_utils import default_weight_loader from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestModelOptNvfp4(CustomTestCase): diff --git a/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_dispatch.py b/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_dispatch.py index e144771556..acb84fbbda 100644 --- a/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_dispatch.py +++ b/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_dispatch.py @@ -12,7 +12,7 @@ tensors, so the tests stay on CPU; the kernels are covered on-device. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from contextlib import contextmanager diff --git a/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_scales.py b/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_scales.py index 4ebb9f5feb..dc266bff4e 100644 --- a/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_scales.py +++ b/test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_scales.py @@ -18,7 +18,7 @@ sizes, so a failure looks like one a real checkpoint would hit. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/layers/quantization/test_mxfp4_flashinfer_activation_prep.py b/test/registered/unit/layers/quantization/test_mxfp4_flashinfer_activation_prep.py index 37b2b1fc8d..5c63c85f97 100644 --- a/test/registered/unit/layers/quantization/test_mxfp4_flashinfer_activation_prep.py +++ b/test/registered/unit/layers/quantization/test_mxfp4_flashinfer_activation_prep.py @@ -12,7 +12,7 @@ from sglang.srt.layers.quantization.mxfp4 import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") per_token_group_quant_module = importlib.import_module( "sglang.kernels.ops.quantization.per_token_group_quant" diff --git a/test/registered/unit/layers/quantization/test_mxfp4_sm100_trtllm_gen.py b/test/registered/unit/layers/quantization/test_mxfp4_sm100_trtllm_gen.py index 60e8f4e361..8644193deb 100644 --- a/test/registered/unit/layers/quantization/test_mxfp4_sm100_trtllm_gen.py +++ b/test/registered/unit/layers/quantization/test_mxfp4_sm100_trtllm_gen.py @@ -20,7 +20,7 @@ from flashinfer import trtllm_fp4_block_scale_moe from sglang.srt.utils import is_sm100_supported from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=120, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=22, stage="base-b", runner_config="4-gpu-b200") if not is_sm100_supported(): pytest.skip( diff --git a/test/registered/unit/layers/quantization/test_mxfp4_sm120_cutlass.py b/test/registered/unit/layers/quantization/test_mxfp4_sm120_cutlass.py index e358e2411a..ad36e1d862 100644 --- a/test/registered/unit/layers/quantization/test_mxfp4_sm120_cutlass.py +++ b/test/registered/unit/layers/quantization/test_mxfp4_sm120_cutlass.py @@ -14,7 +14,7 @@ import torch from sglang.srt.runtime_context import override_platform from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=85, stage="base-b", runner_config="1-gpu-small") @pytest.fixture diff --git a/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py b/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py index 62696f0112..116cba7f4a 100644 --- a/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py +++ b/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py @@ -20,7 +20,7 @@ import torch from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large") +register_cuda_ci(est_time=21, stage="base-b", runner_config="1-gpu-large") flashinfer_fused_moe = pytest.importorskip("flashinfer.fused_moe") diff --git a/test/registered/unit/layers/quantization/test_nvfp4_linear_backends.py b/test/registered/unit/layers/quantization/test_nvfp4_linear_backends.py index 484603e14c..950e88c0a9 100644 --- a/test/registered/unit/layers/quantization/test_nvfp4_linear_backends.py +++ b/test/registered/unit/layers/quantization/test_nvfp4_linear_backends.py @@ -30,7 +30,7 @@ from sglang.test.quant_ref_utils import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=11, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=26, stage="base-b", runner_config="4-gpu-b200") # (M, N, K). The second shape hits the padding paths: N=160 is not a multiple # of 128 (TRTLLM shuffle pad) and K=336 is neither a multiple of 32 (CUTLASS diff --git a/test/registered/unit/layers/quantization/test_nvfp4_moe_backends.py b/test/registered/unit/layers/quantization/test_nvfp4_moe_backends.py index 8ba5f246bc..e0d3ef0a76 100644 --- a/test/registered/unit/layers/quantization/test_nvfp4_moe_backends.py +++ b/test/registered/unit/layers/quantization/test_nvfp4_moe_backends.py @@ -23,7 +23,7 @@ from sglang.test.quant_ref_utils import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=14, stage="base-b", runner_config="4-gpu-b200") +register_cuda_ci(est_time=30, stage="base-b", runner_config="4-gpu-b200") E, H, I, TOPK, M = 8, 1024, 1024, 2, 32 diff --git a/test/registered/unit/layers/quantization/test_quark_config.py b/test/registered/unit/layers/quantization/test_quark_config.py index 42981a1232..ee53d22408 100644 --- a/test/registered/unit/layers/quantization/test_quark_config.py +++ b/test/registered/unit/layers/quantization/test_quark_config.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import sys import types diff --git a/test/registered/unit/layers/quantization/test_quark_utils.py b/test/registered/unit/layers/quantization/test_quark_utils.py index b8dab0d4c7..dc46ad72a9 100644 --- a/test/registered/unit/layers/quantization/test_quark_utils.py +++ b/test/registered/unit/layers/quantization/test_quark_utils.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/layers/quantization/test_unquant_apply_with_addend.py b/test/registered/unit/layers/quantization/test_unquant_apply_with_addend.py index 9651d663a1..6fa944000b 100644 --- a/test/registered/unit/layers/quantization/test_unquant_apply_with_addend.py +++ b/test/registered/unit/layers/quantization/test_unquant_apply_with_addend.py @@ -8,7 +8,7 @@ and leave the caller's buffer intact. from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=25, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-small") import unittest from contextlib import ExitStack diff --git a/test/registered/unit/layers/test_bailing_mrope_shape.py b/test/registered/unit/layers/test_bailing_mrope_shape.py index 862dd5a7a8..0937ed1642 100644 --- a/test/registered/unit/layers/test_bailing_mrope_shape.py +++ b/test/registered/unit/layers/test_bailing_mrope_shape.py @@ -11,7 +11,7 @@ from sglang.srt.layers.rotary_embedding.bailing_mrope import BailingMRotaryEmbed from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _position_config(max_position_embeddings=131072): diff --git a/test/registered/unit/layers/test_conv_layer.py b/test/registered/unit/layers/test_conv_layer.py index 7b5114dd59..4af8807ccc 100644 --- a/test/registered/unit/layers/test_conv_layer.py +++ b/test/registered/unit/layers/test_conv_layer.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/layers/test_dsv41_candidate_blocks.py b/test/registered/unit/layers/test_dsv41_candidate_blocks.py index 79915cf5c2..bba4aaa303 100644 --- a/test/registered/unit/layers/test_dsv41_candidate_blocks.py +++ b/test/registered/unit/layers/test_dsv41_candidate_blocks.py @@ -11,7 +11,7 @@ from sglang.srt.layers.attention.dsv4.candidate_indexer import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestPrefillCandidateBlocks(CustomTestCase): diff --git a/test/registered/unit/layers/test_dsv4_kv_splits_heuristic.py b/test/registered/unit/layers/test_dsv4_kv_splits_heuristic.py index 9f0d5dad54..4dc70d4fe2 100644 --- a/test/registered/unit/layers/test_dsv4_kv_splits_heuristic.py +++ b/test/registered/unit/layers/test_dsv4_kv_splits_heuristic.py @@ -21,7 +21,7 @@ from sglang.kernels.ops.attention.dsv4.unified_kv_kernels.paged_decode import ( ) from sglang.test.ci.ci_register import register_cpu_ci # noqa: E402 -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") # MI355X. Passed explicitly so the tests are device-independent and run on CPU. NUM_CU = 256 diff --git a/test/registered/unit/layers/test_dsv4_nonpaged_indexer.py b/test/registered/unit/layers/test_dsv4_nonpaged_indexer.py index ad06b05e84..3d0552eff2 100644 --- a/test/registered/unit/layers/test_dsv4_nonpaged_indexer.py +++ b/test/registered/unit/layers/test_dsv4_nonpaged_indexer.py @@ -31,7 +31,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=17, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _INDEXER = "sglang.srt.layers.attention.dsv4.indexer" _METADATA = "sglang.srt.layers.attention.dsv4.metadata" diff --git a/test/registered/unit/layers/test_flashattention_paged_mha.py b/test/registered/unit/layers/test_flashattention_paged_mha.py index 7a1799be56..16555cd9db 100644 --- a/test/registered/unit/layers/test_flashattention_paged_mha.py +++ b/test/registered/unit/layers/test_flashattention_paged_mha.py @@ -22,7 +22,7 @@ with patch.dict( FlashAttentionBackend, ) -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _backend(): diff --git a/test/registered/unit/layers/test_flashinfer_comm_fusion.py b/test/registered/unit/layers/test_flashinfer_comm_fusion.py index ac9189a092..ada9c0247b 100644 --- a/test/registered/unit/layers/test_flashinfer_comm_fusion.py +++ b/test/registered/unit/layers/test_flashinfer_comm_fusion.py @@ -11,7 +11,7 @@ from sglang.test.test_utils import CustomTestCase # Collectives are mocked and world_size is a plain int, so the world_size=4 # cases need one real CUDA device. -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") class _FakeWorkspace: diff --git a/test/registered/unit/layers/test_fp8_bpreshuffle_scale.py b/test/registered/unit/layers/test_fp8_bpreshuffle_scale.py index c94e6a4ff9..db721cf5c7 100644 --- a/test/registered/unit/layers/test_fp8_bpreshuffle_scale.py +++ b/test/registered/unit/layers/test_fp8_bpreshuffle_scale.py @@ -13,7 +13,7 @@ from sglang.srt.layers.quantization.fp8_utils import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _simulate_transpose_scale_emit(values: torch.Tensor) -> torch.Tensor: diff --git a/test/registered/unit/layers/test_gdn_mis_metadata.py b/test/registered/unit/layers/test_gdn_mis_metadata.py index 653de57b1f..ebee7bec0f 100644 --- a/test/registered/unit/layers/test_gdn_mis_metadata.py +++ b/test/registered/unit/layers/test_gdn_mis_metadata.py @@ -7,7 +7,7 @@ from sglang.srt.layers.attention.linear.gdn_backend import build_gdn_mis_metadat from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestGDNMISMetadata(CustomTestCase): diff --git a/test/registered/unit/layers/test_kpool_planner_cpu_mirror.py b/test/registered/unit/layers/test_kpool_planner_cpu_mirror.py index 2cf449750a..a95b7efcbb 100644 --- a/test/registered/unit/layers/test_kpool_planner_cpu_mirror.py +++ b/test/registered/unit/layers/test_kpool_planner_cpu_mirror.py @@ -11,7 +11,7 @@ from sglang.srt.layers.attention.dsa import kpool_plan from sglang.srt.layers.attention.dsa.dsa_topk_backend import TopkTransformMethod from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") class _NoDeviceReadTensor(torch.Tensor): diff --git a/test/registered/unit/layers/test_kpool_stream_scheduling.py b/test/registered/unit/layers/test_kpool_stream_scheduling.py index f43c5e6a4b..7c8cd46a41 100644 --- a/test/registered/unit/layers/test_kpool_stream_scheduling.py +++ b/test/registered/unit/layers/test_kpool_stream_scheduling.py @@ -14,7 +14,7 @@ from sglang.srt.layers.attention.dsa.dsa_indexer_kpool import IndexerKPool from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _eager(method): diff --git a/test/registered/unit/layers/test_logprob_chunk_stitching.py b/test/registered/unit/layers/test_logprob_chunk_stitching.py index 51e5cafd98..7c980e3f1d 100644 --- a/test/registered/unit/layers/test_logprob_chunk_stitching.py +++ b/test/registered/unit/layers/test_logprob_chunk_stitching.py @@ -18,7 +18,7 @@ from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.logprob_test_utils import coverage_cases from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") VOCAB = 11 # Heterogeneous per-sequence parameters; uniform ones hide misalignment. diff --git a/test/registered/unit/layers/test_logprob_fast_input.py b/test/registered/unit/layers/test_logprob_fast_input.py index ce72b759b9..0bc15cf599 100644 --- a/test/registered/unit/layers/test_logprob_fast_input.py +++ b/test/registered/unit/layers/test_logprob_fast_input.py @@ -20,7 +20,7 @@ from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.logprob_test_utils import coverage_cases from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") VOCAB = 11 # Heterogeneous per-sequence parameters; uniform ones hide misalignment. diff --git a/test/registered/unit/layers/test_mamba2_track_ssm_indices.py b/test/registered/unit/layers/test_mamba2_track_ssm_indices.py index 48828418f3..f4ab5711ec 100644 --- a/test/registered/unit/layers/test_mamba2_track_ssm_indices.py +++ b/test/registered/unit/layers/test_mamba2_track_ssm_indices.py @@ -11,7 +11,7 @@ import torch from sglang.srt.layers.attention.hybrid_linear_attn_backend import Mamba2AttnBackend from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") CHUNK = 128 diff --git a/test/registered/unit/layers/test_mamba_prefill_track_metadata.py b/test/registered/unit/layers/test_mamba_prefill_track_metadata.py index 397db24288..05e3f5254f 100644 --- a/test/registered/unit/layers/test_mamba_prefill_track_metadata.py +++ b/test/registered/unit/layers/test_mamba_prefill_track_metadata.py @@ -19,7 +19,7 @@ from sglang.srt.runtime_context import get_context from sglang.srt.speculative import spec_utils from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") class NoHostRead(torch.Tensor): diff --git a/test/registered/unit/layers/test_mamba_state_scatter_triton.py b/test/registered/unit/layers/test_mamba_state_scatter_triton.py index 145f20fe7c..06967ed61f 100644 --- a/test/registered/unit/layers/test_mamba_state_scatter_triton.py +++ b/test/registered/unit/layers/test_mamba_state_scatter_triton.py @@ -5,11 +5,11 @@ from sglang.test.ci.ci_register import ( register_xpu_ci, ) -register_cuda_ci(est_time=7, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=7, suite="stage-b-test-1-gpu-small-amd-mi35x") register_xpu_ci(est_time=20, suite="stage-b-test-1-gpu-xpu") # The dst layout-contract tests run on CPU (no kernel launch). -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/layers/test_minicpm_attention_adapter.py b/test/registered/unit/layers/test_minicpm_attention_adapter.py index e08568c0a9..640a99a102 100644 --- a/test/registered/unit/layers/test_minicpm_attention_adapter.py +++ b/test/registered/unit/layers/test_minicpm_attention_adapter.py @@ -24,7 +24,7 @@ with patch.dict( MiniCPMFlashInferAdapter, ) -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") def _metadata(rows=1): diff --git a/test/registered/unit/layers/test_minicpm_sparse_cache.py b/test/registered/unit/layers/test_minicpm_sparse_cache.py index 5369270a1c..c0155c2deb 100644 --- a/test/registered/unit/layers/test_minicpm_sparse_cache.py +++ b/test/registered/unit/layers/test_minicpm_sparse_cache.py @@ -21,7 +21,7 @@ from sglang.srt.server_args import ServerArgs from sglang.srt.session.streaming_session import SessionSlot, StreamingSession from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class RecordingAllocator(BaseTokenToKVPoolAllocator): diff --git a/test/registered/unit/layers/test_minicpm_sparse_metadata.py b/test/registered/unit/layers/test_minicpm_sparse_metadata.py index d11dca84f4..27f0a1de81 100644 --- a/test/registered/unit/layers/test_minicpm_sparse_metadata.py +++ b/test/registered/unit/layers/test_minicpm_sparse_metadata.py @@ -37,7 +37,7 @@ with patch.dict( ) from sglang.srt.runtime_context import get_context, get_schedule -register_cpu_ci(est_time=22, suite="base-a-test-cpu") +register_cpu_ci(est_time=15, suite="base-a-test-cpu") def _compression_layout(): diff --git a/test/registered/unit/layers/test_mova.py b/test/registered/unit/layers/test_mova.py index 72d6b35848..b434366217 100644 --- a/test/registered/unit/layers/test_mova.py +++ b/test/registered/unit/layers/test_mova.py @@ -21,7 +21,7 @@ from sglang.srt.layers.mova import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=4, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") def test_router_bias_changes_selection_but_not_mixture_weights(): diff --git a/test/registered/unit/layers/test_pooler_score_and_pool.py b/test/registered/unit/layers/test_pooler_score_and_pool.py index 4b2ce14a4b..eeddc00b60 100644 --- a/test/registered/unit/layers/test_pooler_score_and_pool.py +++ b/test/registered/unit/layers/test_pooler_score_and_pool.py @@ -19,7 +19,7 @@ from sglang.srt.layers.pooler import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_forward_batch( diff --git a/test/registered/unit/layers/test_radix_attention.py b/test/registered/unit/layers/test_radix_attention.py index 450b830869..dd7b524f88 100644 --- a/test/registered/unit/layers/test_radix_attention.py +++ b/test/registered/unit/layers/test_radix_attention.py @@ -13,7 +13,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _RecordingAttentionBackend: diff --git a/test/registered/unit/layers/test_radix_linear_attention.py b/test/registered/unit/layers/test_radix_linear_attention.py index a11230a8e7..861a809379 100644 --- a/test/registered/unit/layers/test_radix_linear_attention.py +++ b/test/registered/unit/layers/test_radix_linear_attention.py @@ -9,7 +9,7 @@ import sglang.srt.layers.radix_linear_attention as radix_linear_attention from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeAttentionBackend: diff --git a/test/registered/unit/layers/test_yarn_cache_extension.py b/test/registered/unit/layers/test_yarn_cache_extension.py index 361c22f8d2..81858252f3 100644 --- a/test/registered/unit/layers/test_yarn_cache_extension.py +++ b/test/registered/unit/layers/test_yarn_cache_extension.py @@ -7,7 +7,7 @@ from torch.testing import assert_close from sglang.srt.layers.rotary_embedding import base, factory, rope_variant from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") @pytest.mark.parametrize("kind", ("yarn", "deepseek_yarn", "mrope")) diff --git a/test/registered/unit/lora/test_eviction_policy.py b/test/registered/unit/lora/test_eviction_policy.py index a70e87c1c5..d37c2a675e 100644 --- a/test/registered/unit/lora/test_eviction_policy.py +++ b/test/registered/unit/lora/test_eviction_policy.py @@ -20,7 +20,7 @@ from sglang.srt.lora.eviction_policy import get_eviction_policy from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestLoRAEvictionPolicy(CustomTestCase): diff --git a/test/registered/unit/lora/test_experimental_sgl_marlin_runtime_unit.py b/test/registered/unit/lora/test_experimental_sgl_marlin_runtime_unit.py index 5cd7a3bd0d..5c99309971 100644 --- a/test/registered/unit/lora/test_experimental_sgl_marlin_runtime_unit.py +++ b/test/registered/unit/lora/test_experimental_sgl_marlin_runtime_unit.py @@ -19,7 +19,7 @@ from sglang.srt.lora.marlin_lora_temp import moe_runner from sglang.srt.lora.trtllm_lora_temp import environ as trtllm_lora_environ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") _SGLANG_ROOT = Path(sglang.__file__).resolve().parent diff --git a/test/registered/unit/lora/test_inkling_linearized_lora_unit.py b/test/registered/unit/lora/test_inkling_linearized_lora_unit.py index c69de1a163..3b6e67d52c 100644 --- a/test/registered/unit/lora/test_inkling_linearized_lora_unit.py +++ b/test/registered/unit/lora/test_inkling_linearized_lora_unit.py @@ -20,7 +20,7 @@ from sglang.srt.lora.lora_registry import LoRARef from sglang.test.ci.ci_register import register_cpu_ci # Pure layout / bookkeeping math: no CUDA, no distributed groups, no kernels. -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") _HIDDEN = 3 _ADAPTER_RANK = 2 diff --git a/test/registered/unit/lora/test_lm_head_pruning.py b/test/registered/unit/lora/test_lm_head_pruning.py index e3a052f15c..bcaf305516 100644 --- a/test/registered/unit/lora/test_lm_head_pruning.py +++ b/test/registered/unit/lora/test_lm_head_pruning.py @@ -25,7 +25,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestLMHeadPruning(CustomTestCase): diff --git a/test/registered/unit/lora/test_lora_moe_inplace_unit.py b/test/registered/unit/lora/test_lora_moe_inplace_unit.py index b8ce80a823..0dfbd85a58 100644 --- a/test/registered/unit/lora/test_lora_moe_inplace_unit.py +++ b/test/registered/unit/lora/test_lora_moe_inplace_unit.py @@ -24,7 +24,7 @@ Usage: from sglang.test.ci.ci_register import register_cpu_ci # CPU-only unit test; no CUDA/distributed dependencies. -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") import types import unittest diff --git a/test/registered/unit/lora/test_lora_spec_verify_batch_info.py b/test/registered/unit/lora/test_lora_spec_verify_batch_info.py index d7984eeaaa..e74b3d976b 100644 --- a/test/registered/unit/lora/test_lora_spec_verify_batch_info.py +++ b/test/registered/unit/lora/test_lora_spec_verify_batch_info.py @@ -24,7 +24,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small") def _verify_batch(bs: int, draft_token_num: int) -> SimpleNamespace: diff --git a/test/registered/unit/lora/test_mem_pool_ep_unit.py b/test/registered/unit/lora/test_mem_pool_ep_unit.py index 4a3ad357e6..affb3bb208 100644 --- a/test/registered/unit/lora/test_mem_pool_ep_unit.py +++ b/test/registered/unit/lora/test_mem_pool_ep_unit.py @@ -16,7 +16,7 @@ Usage: from sglang.test.ci.ci_register import register_cpu_ci # CPU-only unit test; no CUDA/distributed dependencies. -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import ast import types diff --git a/test/registered/unit/managers/scheduler_components/test_dp_attn.py b/test/registered/unit/managers/scheduler_components/test_dp_attn.py index 5696248f65..15b92ab5ea 100644 --- a/test/registered/unit/managers/scheduler_components/test_dp_attn.py +++ b/test/registered/unit/managers/scheduler_components/test_dp_attn.py @@ -12,7 +12,7 @@ from sglang.srt.managers.scheduler_components import dp_attn # noqa: E402 from sglang.srt.model_executor.forward_batch_info import ForwardMode # noqa: E402 from sglang.srt.speculative.spec_info import SpeculativeAlgorithm # noqa: E402 -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=73, suite="base-a-test-cpu") class TestDPAttnSchedulerMetadata(CustomTestCase): diff --git a/test/registered/unit/managers/scheduler_components/test_invariant_checker.py b/test/registered/unit/managers/scheduler_components/test_invariant_checker.py index 97eeb06f67..8091c1b96a 100644 --- a/test/registered/unit/managers/scheduler_components/test_invariant_checker.py +++ b/test/registered/unit/managers/scheduler_components/test_invariant_checker.py @@ -14,7 +14,7 @@ from sglang.srt.managers.scheduler_components.invariant_checker import ( SchedulerInvariantChecker, ) -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestCheckTreeCacheGate(CustomTestCase): diff --git a/test/registered/unit/managers/scheduler_components/test_output_sender.py b/test/registered/unit/managers/scheduler_components/test_output_sender.py index 85110137c6..87101825bb 100644 --- a/test/registered/unit/managers/scheduler_components/test_output_sender.py +++ b/test/registered/unit/managers/scheduler_components/test_output_sender.py @@ -13,7 +13,7 @@ from sglang.srt.managers.schedule_batch import Req from sglang.srt.managers.scheduler_components.output_sender import SenderWrapper from sglang.srt.sampling.sampling_params import SamplingParams -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_scheduler_req(http_worker_ipc: str) -> Req: diff --git a/test/registered/unit/managers/test_auxiliary_output.py b/test/registered/unit/managers/test_auxiliary_output.py index 4265ea61ee..e3368a189e 100644 --- a/test/registered/unit/managers/test_auxiliary_output.py +++ b/test/registered/unit/managers/test_auxiliary_output.py @@ -30,7 +30,7 @@ from sglang.srt.speculative.spec_info import SpeculativeAlgorithm from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import published_topology -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") @dataclass diff --git a/test/registered/unit/managers/test_bailing_modality_metadata.py b/test/registered/unit/managers/test_bailing_modality_metadata.py index 2a75cad748..c257604cc1 100644 --- a/test/registered/unit/managers/test_bailing_modality_metadata.py +++ b/test/registered/unit/managers/test_bailing_modality_metadata.py @@ -21,7 +21,7 @@ from sglang.srt.model_executor.forward_batch_info import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=15, suite="base-a-test-cpu") +register_cpu_ci(est_time=20, suite="base-a-test-cpu") class TestBailingModalityMetadata(CustomTestCase): diff --git a/test/registered/unit/managers/test_batch_result_processor_hidden_states.py b/test/registered/unit/managers/test_batch_result_processor_hidden_states.py index 363515e103..ac227dbde5 100644 --- a/test/registered/unit/managers/test_batch_result_processor_hidden_states.py +++ b/test/registered/unit/managers/test_batch_result_processor_hidden_states.py @@ -15,7 +15,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_processor(case, server_mode: str = "full") -> SchedulerBatchResultProcessor: diff --git a/test/registered/unit/managers/test_batch_result_processor_mamba_boundary.py b/test/registered/unit/managers/test_batch_result_processor_mamba_boundary.py index 97eb071fb1..17d8e23f76 100644 --- a/test/registered/unit/managers/test_batch_result_processor_mamba_boundary.py +++ b/test/registered/unit/managers/test_batch_result_processor_mamba_boundary.py @@ -15,7 +15,7 @@ from sglang.srt.runtime_context import get_context from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # The decode checkpoint grid is lcm(mamba_cache_chunk_size, tree page, diff --git a/test/registered/unit/managers/test_batch_result_processor_spec_grammar.py b/test/registered/unit/managers/test_batch_result_processor_spec_grammar.py index 6fc1a88c42..503d0016a3 100644 --- a/test/registered/unit/managers/test_batch_result_processor_spec_grammar.py +++ b/test/registered/unit/managers/test_batch_result_processor_spec_grammar.py @@ -22,7 +22,7 @@ from sglang.srt.sampling.sampling_params import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") class _FakeGrammar: diff --git a/test/registered/unit/managers/test_data_parallel_controller.py b/test/registered/unit/managers/test_data_parallel_controller.py index c17ad1f712..65ee2b6f64 100644 --- a/test/registered/unit/managers/test_data_parallel_controller.py +++ b/test/registered/unit/managers/test_data_parallel_controller.py @@ -30,7 +30,7 @@ from sglang.srt.managers.data_parallel_controller import ( ) from sglang.srt.managers.load_snapshot import LoadSnapshot -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") _BASE_LOAD = msgspec.structs.replace( diff --git a/test/registered/unit/managers/test_detailed_annotations.py b/test/registered/unit/managers/test_detailed_annotations.py index 604af818b7..5eee8ec7ec 100644 --- a/test/registered/unit/managers/test_detailed_annotations.py +++ b/test/registered/unit/managers/test_detailed_annotations.py @@ -25,7 +25,7 @@ from sglang.srt.model_executor.step_span_utils import ( ) from sglang.srt.utils.profile_utils import build_step_span_name -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _CpuMirror: diff --git a/test/registered/unit/managers/test_disagg_idle_step_counters.py b/test/registered/unit/managers/test_disagg_idle_step_counters.py index 90a769d534..313789a82e 100644 --- a/test/registered/unit/managers/test_disagg_idle_step_counters.py +++ b/test/registered/unit/managers/test_disagg_idle_step_counters.py @@ -26,7 +26,7 @@ from sglang.test.test_utils import ( published_topology, ) -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") LAUNCH_TIMESTAMPS = (0.0, 0.125, 1.0, 1.125) # Prefill busy time is charged launch -> result, so the result clock matters too. diff --git a/test/registered/unit/managers/test_embed_overrides.py b/test/registered/unit/managers/test_embed_overrides.py index e00a1d1fc4..07308effc2 100644 --- a/test/registered/unit/managers/test_embed_overrides.py +++ b/test/registered/unit/managers/test_embed_overrides.py @@ -26,7 +26,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") HIDDEN_DIM = 4 diff --git a/test/registered/unit/managers/test_fanout_communicator.py b/test/registered/unit/managers/test_fanout_communicator.py index d40687af6e..c129758ee8 100644 --- a/test/registered/unit/managers/test_fanout_communicator.py +++ b/test/registered/unit/managers/test_fanout_communicator.py @@ -7,7 +7,7 @@ from sglang.srt.managers.communicator import FanOutCommunicator from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestQueueingCall(CustomTestCase): diff --git a/test/registered/unit/managers/test_finish_length_speculative.py b/test/registered/unit/managers/test_finish_length_speculative.py index 5e1a174de6..7d5fee5d68 100644 --- a/test/registered/unit/managers/test_finish_length_speculative.py +++ b/test/registered/unit/managers/test_finish_length_speculative.py @@ -25,7 +25,7 @@ from sglang.srt.managers.schedule_batch import ( ) from sglang.srt.sampling.sampling_params import SamplingParams -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") EOS_ID = 2 STOP_ID = 1 diff --git a/test/registered/unit/managers/test_flat_raw_top_logprobs.py b/test/registered/unit/managers/test_flat_raw_top_logprobs.py index f7e4e09e64..ce59189ed1 100644 --- a/test/registered/unit/managers/test_flat_raw_top_logprobs.py +++ b/test/registered/unit/managers/test_flat_raw_top_logprobs.py @@ -40,7 +40,7 @@ from sglang.srt.managers.tokenizer_manager import ( from sglang.srt.observability.req_time_stats import APIServerReqTimeStats from sglang.srt.sampling.sampling_params import SamplingParams -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # Synthetic per-position top-k rows (k=2). The leading None mirrors the first # prompt position, which has no top logprobs. diff --git a/test/registered/unit/managers/test_grammar_stop_speculative.py b/test/registered/unit/managers/test_grammar_stop_speculative.py index 0a60c2b8f0..74d5d0ea02 100644 --- a/test/registered/unit/managers/test_grammar_stop_speculative.py +++ b/test/registered/unit/managers/test_grammar_stop_speculative.py @@ -7,7 +7,7 @@ from sglang.srt.managers.schedule_batch import Req from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") EOS_TOKEN_ID = 2 STOP_TOKEN_ID = 17 diff --git a/test/registered/unit/managers/test_hidden_state_server_mode.py b/test/registered/unit/managers/test_hidden_state_server_mode.py index 5aa30c20a8..73f82d5a97 100644 --- a/test/registered/unit/managers/test_hidden_state_server_mode.py +++ b/test/registered/unit/managers/test_hidden_state_server_mode.py @@ -8,7 +8,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHiddenStateServerMode(CustomTestCase): diff --git a/test/registered/unit/managers/test_hisparse_unit.py b/test/registered/unit/managers/test_hisparse_unit.py index 459c100054..7bee1c34f0 100644 --- a/test/registered/unit/managers/test_hisparse_unit.py +++ b/test/registered/unit/managers/test_hisparse_unit.py @@ -31,7 +31,7 @@ from sglang.test.ci.ci_register import ( register_xpu_ci, ) -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd") register_xpu_ci(est_time=60, suite="stage-b-test-1-gpu-xpu") diff --git a/test/registered/unit/managers/test_kv_page_invariants.py b/test/registered/unit/managers/test_kv_page_invariants.py index 9ba86262b2..3f18b9ea4d 100644 --- a/test/registered/unit/managers/test_kv_page_invariants.py +++ b/test/registered/unit/managers/test_kv_page_invariants.py @@ -9,7 +9,7 @@ from sglang.srt.managers.schedule_batch import ReqKvInfo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _PAGE_SIZE = 256 diff --git a/test/registered/unit/managers/test_load_inquirer.py b/test/registered/unit/managers/test_load_inquirer.py index 7fdfdef610..6ec37c3961 100644 --- a/test/registered/unit/managers/test_load_inquirer.py +++ b/test/registered/unit/managers/test_load_inquirer.py @@ -12,7 +12,7 @@ from sglang.srt.managers.scheduler_components.load_inquirer import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestSchedulePolicyWaitingQueueMatching(unittest.TestCase): diff --git a/test/registered/unit/managers/test_load_snapshot_backends.py b/test/registered/unit/managers/test_load_snapshot_backends.py index e93bbfcb8f..efe0e4870f 100644 --- a/test/registered/unit/managers/test_load_snapshot_backends.py +++ b/test/registered/unit/managers/test_load_snapshot_backends.py @@ -26,7 +26,7 @@ from sglang.test.test_utils import CustomTestCase, maybe_stub_sgl_kernel maybe_stub_sgl_kernel() -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _temp_path() -> str: diff --git a/test/registered/unit/managers/test_loadstat_wire.py b/test/registered/unit/managers/test_loadstat_wire.py index 02a55f6643..e0bca4ac49 100644 --- a/test/registered/unit/managers/test_loadstat_wire.py +++ b/test/registered/unit/managers/test_loadstat_wire.py @@ -30,7 +30,7 @@ from sglang.srt.managers.scheduler_components.load_publisher import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase, published_topology -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestLoadStatWire(CustomTestCase): diff --git a/test/registered/unit/managers/test_lora_update_result_merge.py b/test/registered/unit/managers/test_lora_update_result_merge.py index c91b24a6d4..fb4cc221b8 100644 --- a/test/registered/unit/managers/test_lora_update_result_merge.py +++ b/test/registered/unit/managers/test_lora_update_result_merge.py @@ -21,7 +21,7 @@ from sglang.srt.managers.tokenizer_control_mixin import _merge_lora_update_resul from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _ok(adapters=None) -> LoRAUpdateOutput: diff --git a/test/registered/unit/managers/test_mamba_checkpoint_depth.py b/test/registered/unit/managers/test_mamba_checkpoint_depth.py index dd4e62c03b..ba8b7fea76 100644 --- a/test/registered/unit/managers/test_mamba_checkpoint_depth.py +++ b/test/registered/unit/managers/test_mamba_checkpoint_depth.py @@ -25,7 +25,7 @@ from sglang.srt.server_args import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") CHUNK = 64 diff --git a/test/registered/unit/managers/test_mm_hashes.py b/test/registered/unit/managers/test_mm_hashes.py index c35dca3623..49c844ce75 100644 --- a/test/registered/unit/managers/test_mm_hashes.py +++ b/test/registered/unit/managers/test_mm_hashes.py @@ -25,7 +25,7 @@ from sglang.srt.managers.schedule_batch import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestMmHashesContract(CustomTestCase): diff --git a/test/registered/unit/managers/test_mm_process_config.py b/test/registered/unit/managers/test_mm_process_config.py index e9c8ea3d62..df35e7d365 100644 --- a/test/registered/unit/managers/test_mm_process_config.py +++ b/test/registered/unit/managers/test_mm_process_config.py @@ -14,7 +14,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestMmProcessConfigValidation(CustomTestCase): diff --git a/test/registered/unit/managers/test_mm_shm_error_consensus.py b/test/registered/unit/managers/test_mm_shm_error_consensus.py index a7d1e20f5d..14227d5c94 100644 --- a/test/registered/unit/managers/test_mm_shm_error_consensus.py +++ b/test/registered/unit/managers/test_mm_shm_error_consensus.py @@ -42,7 +42,7 @@ from sglang.srt.managers.scheduler_components.request_receiver import ( # noqa: ) from sglang.srt.sampling.sampling_params import SamplingParams # noqa: E402 -register_cpu_ci(est_time=33, suite="base-a-test-cpu") +register_cpu_ci(est_time=67, suite="base-a-test-cpu") class _CloneFailure: diff --git a/test/registered/unit/managers/test_mm_utils_split.py b/test/registered/unit/managers/test_mm_utils_split.py index 59bc876380..1d9f0f9e37 100644 --- a/test/registered/unit/managers/test_mm_utils_split.py +++ b/test/registered/unit/managers/test_mm_utils_split.py @@ -21,7 +21,7 @@ from sglang.srt.managers.schedule_batch import Modality, MultimodalDataItem from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _bundled_item(grid_key=None, grid=None, feature_len=10, num_images=2): diff --git a/test/registered/unit/managers/test_multi_tokenizer_mixin.py b/test/registered/unit/managers/test_multi_tokenizer_mixin.py index 7468d1fc50..699a0563f6 100644 --- a/test/registered/unit/managers/test_multi_tokenizer_mixin.py +++ b/test/registered/unit/managers/test_multi_tokenizer_mixin.py @@ -13,7 +13,7 @@ from sglang.srt.managers.multi_tokenizer_mixin import ( get_tokenizer_worker_class, ) -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class CustomTokenizerWorker(TokenizerWorker): diff --git a/test/registered/unit/managers/test_multimodal_abort_cleanup.py b/test/registered/unit/managers/test_multimodal_abort_cleanup.py index 8ddc42b2a5..24e70cab6c 100644 --- a/test/registered/unit/managers/test_multimodal_abort_cleanup.py +++ b/test/registered/unit/managers/test_multimodal_abort_cleanup.py @@ -13,7 +13,7 @@ from sglang.srt.managers.schedule_batch import ( from sglang.srt.multimodal.transport.cuda_ipc import CudaIpcTensorTransportProxy from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _deferred_proxy(): diff --git a/test/registered/unit/managers/test_output_streamer_customized_info.py b/test/registered/unit/managers/test_output_streamer_customized_info.py index d6b438f2ee..ec9785e37d 100644 --- a/test/registered/unit/managers/test_output_streamer_customized_info.py +++ b/test/registered/unit/managers/test_output_streamer_customized_info.py @@ -16,7 +16,7 @@ from sglang.srt.utils.weight_versions import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import enter_scope, published_topology -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeReq: diff --git a/test/registered/unit/managers/test_output_streamer_logprobs.py b/test/registered/unit/managers/test_output_streamer_logprobs.py index 30b79283b7..d2d53edb10 100644 --- a/test/registered/unit/managers/test_output_streamer_logprobs.py +++ b/test/registered/unit/managers/test_output_streamer_logprobs.py @@ -8,7 +8,7 @@ from sglang.srt.managers.scheduler_components.output_streamer import ( from sglang.srt.speculative.spec_info import SpeculativeAlgorithm from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeReq: diff --git a/test/registered/unit/managers/test_pp_cp_rank_offsets.py b/test/registered/unit/managers/test_pp_cp_rank_offsets.py index 5db7cd72bc..17bf59654c 100644 --- a/test/registered/unit/managers/test_pp_cp_rank_offsets.py +++ b/test/registered/unit/managers/test_pp_cp_rank_offsets.py @@ -23,7 +23,7 @@ from sglang.srt.managers.scheduler_pp_mixin import ( # noqa: E402 ) from sglang.srt.model_executor.forward_batch_info import ForwardMode # noqa: E402 -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _published_topology(): diff --git a/test/registered/unit/managers/test_prefill_adder.py b/test/registered/unit/managers/test_prefill_adder.py index 03b21a95a0..a90c09b31a 100644 --- a/test/registered/unit/managers/test_prefill_adder.py +++ b/test/registered/unit/managers/test_prefill_adder.py @@ -30,7 +30,7 @@ from sglang.srt.utils.common import Range from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") class _RecordingDelayer: diff --git a/test/registered/unit/managers/test_prefill_delayer_high_watermark.py b/test/registered/unit/managers/test_prefill_delayer_high_watermark.py index 2631f4e5ba..8db3d500f8 100644 --- a/test/registered/unit/managers/test_prefill_delayer_high_watermark.py +++ b/test/registered/unit/managers/test_prefill_delayer_high_watermark.py @@ -9,7 +9,7 @@ from sglang.srt.managers.prefill_delayer import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestPrefillDelayerHighWatermark(CustomTestCase): diff --git a/test/registered/unit/managers/test_priority_scheduling_disaggregation.py b/test/registered/unit/managers/test_priority_scheduling_disaggregation.py index d8a3e4cbc7..231e2b2738 100644 --- a/test/registered/unit/managers/test_priority_scheduling_disaggregation.py +++ b/test/registered/unit/managers/test_priority_scheduling_disaggregation.py @@ -28,7 +28,7 @@ from sglang.test.separate_buffer_allocator_double import ( bind_separate_buffer_capacity, ) -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDisaggregationPriorityQueueing(unittest.TestCase): diff --git a/test/registered/unit/managers/test_retraction_order.py b/test/registered/unit/managers/test_retraction_order.py index abd8318382..3e54473e45 100644 --- a/test/registered/unit/managers/test_retraction_order.py +++ b/test/registered/unit/managers/test_retraction_order.py @@ -7,7 +7,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _req(output_len: int, input_len: int = 8, priority=None): diff --git a/test/registered/unit/managers/test_sampling_mask_validation.py b/test/registered/unit/managers/test_sampling_mask_validation.py index a41dfe5485..a7a4a72abf 100644 --- a/test/registered/unit/managers/test_sampling_mask_validation.py +++ b/test/registered/unit/managers/test_sampling_mask_validation.py @@ -12,7 +12,7 @@ from sglang.srt.sampling.custom_logit_processor import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestSamplingMaskValidation(CustomTestCase): diff --git a/test/registered/unit/managers/test_schedule_batch_convert_decode_to_extend.py b/test/registered/unit/managers/test_schedule_batch_convert_decode_to_extend.py index 176cb4cef8..c19d24819e 100644 --- a/test/registered/unit/managers/test_schedule_batch_convert_decode_to_extend.py +++ b/test/registered/unit/managers/test_schedule_batch_convert_decode_to_extend.py @@ -21,7 +21,7 @@ from sglang.srt.managers.schedule_batch import ( # noqa: E402 ScheduleBatch, ) -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeReq: diff --git a/test/registered/unit/managers/test_schedule_batch_out_of_place.py b/test/registered/unit/managers/test_schedule_batch_out_of_place.py index 867ab1ee47..7be87fd8b0 100644 --- a/test/registered/unit/managers/test_schedule_batch_out_of_place.py +++ b/test/registered/unit/managers/test_schedule_batch_out_of_place.py @@ -19,7 +19,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode # noqa: E4 from sglang.srt.speculative.spec_info import SpeculativeAlgorithm # noqa: E402 from sglang.srt.utils.common import Range # noqa: E402 -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") AUTO_FILL_EXCLUDED_FIELDS = ["reqs"] diff --git a/test/registered/unit/managers/test_schedule_batch_prepare_for_decode.py b/test/registered/unit/managers/test_schedule_batch_prepare_for_decode.py index ca47ad7007..14c29dc0f2 100644 --- a/test/registered/unit/managers/test_schedule_batch_prepare_for_decode.py +++ b/test/registered/unit/managers/test_schedule_batch_prepare_for_decode.py @@ -12,7 +12,7 @@ maybe_stub_sgl_kernel() from sglang.srt.managers.schedule_batch import ScheduleBatch # noqa: E402 -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_req(): diff --git a/test/registered/unit/managers/test_schedule_batch_req_pool_indices.py b/test/registered/unit/managers/test_schedule_batch_req_pool_indices.py index 3fc07caf22..052d66092f 100644 --- a/test/registered/unit/managers/test_schedule_batch_req_pool_indices.py +++ b/test/registered/unit/managers/test_schedule_batch_req_pool_indices.py @@ -13,7 +13,7 @@ from sglang.srt.managers.hisparse_coordinator import HiSparseCoordinator # noqa from sglang.srt.managers.scheduler import Scheduler # noqa: E402 from sglang.srt.model_executor.forward_batch_info import CaptureHiddenMode # noqa: E402 -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_req(req_pool_idx, origin_input_ids, output_ids): diff --git a/test/registered/unit/managers/test_schedule_policy.py b/test/registered/unit/managers/test_schedule_policy.py index 45ce93734a..53a7bb2c57 100644 --- a/test/registered/unit/managers/test_schedule_policy.py +++ b/test/registered/unit/managers/test_schedule_policy.py @@ -9,7 +9,7 @@ from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_req(rid, origin_input_text, origin_input_ids, sampling_params=None, **kwargs): diff --git a/test/registered/unit/managers/test_scheduler_chunked_abort_race.py b/test/registered/unit/managers/test_scheduler_chunked_abort_race.py index b848b69d9b..744d0d392c 100644 --- a/test/registered/unit/managers/test_scheduler_chunked_abort_race.py +++ b/test/registered/unit/managers/test_scheduler_chunked_abort_race.py @@ -16,7 +16,7 @@ maybe_stub_sgl_kernel() from sglang.srt.managers.scheduler import Scheduler # noqa: E402 -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeReq: diff --git a/test/registered/unit/managers/test_scheduler_chunked_req_gate.py b/test/registered/unit/managers/test_scheduler_chunked_req_gate.py index 94ac81dd53..94a02281cb 100644 --- a/test/registered/unit/managers/test_scheduler_chunked_req_gate.py +++ b/test/registered/unit/managers/test_scheduler_chunked_req_gate.py @@ -18,7 +18,7 @@ from sglang.srt.managers.scheduler import Scheduler from sglang.srt.mem_cache.chunk_cache import ChunkCache from sglang.srt.utils.common import Range -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_req( diff --git a/test/registered/unit/managers/test_scheduler_flush_cache.py b/test/registered/unit/managers/test_scheduler_flush_cache.py index a5030fcd46..e923761a55 100644 --- a/test/registered/unit/managers/test_scheduler_flush_cache.py +++ b/test/registered/unit/managers/test_scheduler_flush_cache.py @@ -12,7 +12,7 @@ from sglang.srt.managers.scheduler_components.flush_wrapper import ( SchedulerFlushWrapper, ) -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/managers/test_scheduler_hicache_attach.py b/test/registered/unit/managers/test_scheduler_hicache_attach.py index f5e567b1a5..f9b0b19095 100644 --- a/test/registered/unit/managers/test_scheduler_hicache_attach.py +++ b/test/registered/unit/managers/test_scheduler_hicache_attach.py @@ -19,7 +19,7 @@ from sglang.srt.runtime_context import get_context, get_memory from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestSchedulerHiCacheAttach(CustomTestCase): diff --git a/test/registered/unit/managers/test_scheduler_hicache_events.py b/test/registered/unit/managers/test_scheduler_hicache_events.py index 8c81c07927..b70bcc28ea 100644 --- a/test/registered/unit/managers/test_scheduler_hicache_events.py +++ b/test/registered/unit/managers/test_scheduler_hicache_events.py @@ -16,7 +16,7 @@ maybe_stub_sgl_kernel() from sglang.srt.managers.schedule_batch import NextBatchPlan from sglang.srt.managers.scheduler import Scheduler -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestSchedulerHiCacheEvents(unittest.TestCase): diff --git a/test/registered/unit/managers/test_scheduler_init_req_max_new_tokens.py b/test/registered/unit/managers/test_scheduler_init_req_max_new_tokens.py index d651071e5d..5b36b5155a 100644 --- a/test/registered/unit/managers/test_scheduler_init_req_max_new_tokens.py +++ b/test/registered/unit/managers/test_scheduler_init_req_max_new_tokens.py @@ -11,7 +11,7 @@ from sglang.srt.mem_cache.unified_memory_pool import init_unified_swa_pools from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestSchedulerInitReqMaxNewTokens(unittest.TestCase): diff --git a/test/registered/unit/managers/test_scheduler_on_idle_load.py b/test/registered/unit/managers/test_scheduler_on_idle_load.py index 9cba96e57c..acd4f88415 100644 --- a/test/registered/unit/managers/test_scheduler_on_idle_load.py +++ b/test/registered/unit/managers/test_scheduler_on_idle_load.py @@ -17,7 +17,7 @@ maybe_stub_sgl_kernel() from sglang.srt.disaggregation.utils import DisaggregationMode from sglang.srt.managers.scheduler import Scheduler -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestOnIdleStallPublish(CustomTestCase): diff --git a/test/registered/unit/managers/test_scheduler_pause_generation.py b/test/registered/unit/managers/test_scheduler_pause_generation.py index 716c59e578..205a154bd6 100644 --- a/test/registered/unit/managers/test_scheduler_pause_generation.py +++ b/test/registered/unit/managers/test_scheduler_pause_generation.py @@ -23,7 +23,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.srt.runtime_context import publish, reset_context from sglang.srt.sampling.sampling_params import SamplingParams -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/managers/test_scheduler_prefill_decode_interval.py b/test/registered/unit/managers/test_scheduler_prefill_decode_interval.py index d238916531..ae4fe6db72 100644 --- a/test/registered/unit/managers/test_scheduler_prefill_decode_interval.py +++ b/test/registered/unit/managers/test_scheduler_prefill_decode_interval.py @@ -10,7 +10,7 @@ maybe_stub_sgl_kernel() from sglang.srt.managers.scheduler import Scheduler -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_scheduler(*, interval: int, require_mlp_sync: bool) -> Scheduler: diff --git a/test/registered/unit/managers/test_scheduler_recv_skipper.py b/test/registered/unit/managers/test_scheduler_recv_skipper.py index e912ffabef..f5b6feab11 100644 --- a/test/registered/unit/managers/test_scheduler_recv_skipper.py +++ b/test/registered/unit/managers/test_scheduler_recv_skipper.py @@ -12,7 +12,7 @@ from sglang.srt.managers.scheduler_components.recv_skipper import ( # noqa: E40 ) from sglang.srt.model_executor.forward_batch_info import ForwardMode # noqa: E402 -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _publish(case, interval, enable_dp_attention=False): diff --git a/test/registered/unit/managers/test_scheduler_sampling_mask_validation.py b/test/registered/unit/managers/test_scheduler_sampling_mask_validation.py index ebc47b313c..aefcd9c15f 100644 --- a/test/registered/unit/managers/test_scheduler_sampling_mask_validation.py +++ b/test/registered/unit/managers/test_scheduler_sampling_mask_validation.py @@ -13,7 +13,7 @@ maybe_stub_sgl_kernel() from sglang.srt.disaggregation.utils import DisaggregationMode from sglang.srt.managers.scheduler import Scheduler -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestSchedulerSamplingMaskValidation(CustomTestCase): diff --git a/test/registered/unit/managers/test_scheduler_timeouts.py b/test/registered/unit/managers/test_scheduler_timeouts.py index a6c152e32e..df3d44987b 100644 --- a/test/registered/unit/managers/test_scheduler_timeouts.py +++ b/test/registered/unit/managers/test_scheduler_timeouts.py @@ -27,7 +27,7 @@ from sglang.srt.mem_cache.base_prefix_cache import ( CacheRequestOutcome, ) -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeReq: diff --git a/test/registered/unit/managers/test_scheduler_weight_version_tracking.py b/test/registered/unit/managers/test_scheduler_weight_version_tracking.py index a3d1c61da3..d908d2119e 100644 --- a/test/registered/unit/managers/test_scheduler_weight_version_tracking.py +++ b/test/registered/unit/managers/test_scheduler_weight_version_tracking.py @@ -9,7 +9,7 @@ from sglang.srt.managers.scheduler_components.weight_updater import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _ServingStub: diff --git a/test/registered/unit/managers/test_stop_str_speculative.py b/test/registered/unit/managers/test_stop_str_speculative.py index b3c9a26baf..c96dd971a7 100644 --- a/test/registered/unit/managers/test_stop_str_speculative.py +++ b/test/registered/unit/managers/test_stop_str_speculative.py @@ -12,7 +12,7 @@ from sglang.srt.managers.schedule_batch import Req from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # Token id -> decoded text; decode() concatenates. Distinct symbols so no # accidental cross-matches (10-39 are lowercase letters). diff --git a/test/registered/unit/managers/test_tokenizer_config_updates.py b/test/registered/unit/managers/test_tokenizer_config_updates.py index da73e9dcea..e0cbd72ecc 100644 --- a/test/registered/unit/managers/test_tokenizer_config_updates.py +++ b/test/registered/unit/managers/test_tokenizer_config_updates.py @@ -20,7 +20,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _manager(case, **fields): diff --git a/test/registered/unit/managers/test_tokenizer_manager_rid_cleanup.py b/test/registered/unit/managers/test_tokenizer_manager_rid_cleanup.py index 2c09c2efc9..dcab6ed507 100644 --- a/test/registered/unit/managers/test_tokenizer_manager_rid_cleanup.py +++ b/test/registered/unit/managers/test_tokenizer_manager_rid_cleanup.py @@ -38,7 +38,7 @@ from sglang.srt.observability.req_time_stats import ( # noqa: E402 ) from sglang.srt.runtime_context import get_context -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") _NOT_FINISHED = object() # Sentinel: request has not finished yet diff --git a/test/registered/unit/managers/test_trim_matched_stop.py b/test/registered/unit/managers/test_trim_matched_stop.py index 8c5fe6cf40..aeff2f14ed 100644 --- a/test/registered/unit/managers/test_trim_matched_stop.py +++ b/test/registered/unit/managers/test_trim_matched_stop.py @@ -12,7 +12,7 @@ from types import SimpleNamespace from sglang.srt.managers.detokenizer_manager import DetokenizerManager from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") GPT_OSS_CALL_TOKEN = 200012 diff --git a/test/registered/unit/managers/test_vocab_boundary_finish.py b/test/registered/unit/managers/test_vocab_boundary_finish.py index 9c90459a8c..2e6040f52d 100644 --- a/test/registered/unit/managers/test_vocab_boundary_finish.py +++ b/test/registered/unit/managers/test_vocab_boundary_finish.py @@ -9,7 +9,7 @@ maybe_stub_sgl_kernel() from sglang.srt.managers.schedule_batch import FINISH_MATCHED_STR, Req -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") VOCAB_SIZE = 1000 diff --git a/test/registered/unit/mem_cache/test_asymmetric_mha_pool.py b/test/registered/unit/mem_cache/test_asymmetric_mha_pool.py index 6fd07948ce..c3702dbbfb 100644 --- a/test/registered/unit/mem_cache/test_asymmetric_mha_pool.py +++ b/test/registered/unit/mem_cache/test_asymmetric_mha_pool.py @@ -21,7 +21,7 @@ from sglang.test.ci.ci_register import register_cuda_ci _HAS_CUDA = torch.cuda.is_available() -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") DTYPE = torch.bfloat16 HEAD_NUM = 2 diff --git a/test/registered/unit/mem_cache/test_asymmetric_mha_pool_host_unit.py b/test/registered/unit/mem_cache/test_asymmetric_mha_pool_host_unit.py index d6fb6de05f..1175c55c1b 100644 --- a/test/registered/unit/mem_cache/test_asymmetric_mha_pool_host_unit.py +++ b/test/registered/unit/mem_cache/test_asymmetric_mha_pool_host_unit.py @@ -14,7 +14,7 @@ from sglang.srt.mem_cache.pool_host.mha import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_host(layout: str) -> AsymmetricMHATokenToKVPoolHost: diff --git a/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py b/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py index 8f237b6f6b..8da8ff9df2 100644 --- a/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py +++ b/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py @@ -25,7 +25,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import enter_override -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from array import array diff --git a/test/registered/unit/mem_cache/test_decode_retraction_backup.py b/test/registered/unit/mem_cache/test_decode_retraction_backup.py index 2d05894381..a07b41f18b 100644 --- a/test/registered/unit/mem_cache/test_decode_retraction_backup.py +++ b/test/registered/unit/mem_cache/test_decode_retraction_backup.py @@ -24,7 +24,7 @@ from sglang.srt.speculative.base_spec_worker import ( ) from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") class TestDecodeRetractionBackup(unittest.TestCase): diff --git a/test/registered/unit/mem_cache/test_dllm_fdfo_kv_reuse.py b/test/registered/unit/mem_cache/test_dllm_fdfo_kv_reuse.py index 7af5cea53d..0f5168c186 100644 --- a/test/registered/unit/mem_cache/test_dllm_fdfo_kv_reuse.py +++ b/test/registered/unit/mem_cache/test_dllm_fdfo_kv_reuse.py @@ -13,7 +13,7 @@ from sglang.srt.mem_cache.memory_pool import ReqToTokenPool from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeAllocator: diff --git a/test/registered/unit/mem_cache/test_dsa_layer_shard_utils.py b/test/registered/unit/mem_cache/test_dsa_layer_shard_utils.py index 755ac7686c..5f604054ba 100644 --- a/test/registered/unit/mem_cache/test_dsa_layer_shard_utils.py +++ b/test/registered/unit/mem_cache/test_dsa_layer_shard_utils.py @@ -8,7 +8,7 @@ from sglang.srt.mem_cache.dsa_cache_layer_split import LayerSplitDSATokenToKVPoo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestDSALayerShardUtils(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_dsa_layer_split_broadcast.py b/test/registered/unit/mem_cache/test_dsa_layer_split_broadcast.py index 209b7cd499..8f40dc885a 100644 --- a/test/registered/unit/mem_cache/test_dsa_layer_split_broadcast.py +++ b/test/registered/unit/mem_cache/test_dsa_layer_split_broadcast.py @@ -22,7 +22,7 @@ import torch.multiprocessing as mp from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase, publish_build_topology -register_cuda_ci(est_time=28, stage="base-c", runner_config="4-gpu-b200") +register_cuda_ci(est_time=30, stage="base-c", runner_config="4-gpu-b200") LAYER_NUM = 4 PAGE_SIZE = 64 diff --git a/test/registered/unit/mem_cache/test_dsa_pool_host_unit.py b/test/registered/unit/mem_cache/test_dsa_pool_host_unit.py index cb8d5b0f0e..490d7a9291 100644 --- a/test/registered/unit/mem_cache/test_dsa_pool_host_unit.py +++ b/test/registered/unit/mem_cache/test_dsa_pool_host_unit.py @@ -13,7 +13,7 @@ from sglang.srt.mem_cache.pool_host.mla import MLATokenToKVPoolHost from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/unit/mem_cache/test_dsv4_c4_state_lifecycle.py b/test/registered/unit/mem_cache/test_dsv4_c4_state_lifecycle.py index c0b2b7a1a8..3193b87332 100644 --- a/test/registered/unit/mem_cache/test_dsv4_c4_state_lifecycle.py +++ b/test/registered/unit/mem_cache/test_dsv4_c4_state_lifecycle.py @@ -14,7 +14,7 @@ from sglang.srt.mem_cache.memory_pool import ReqToTokenPool from sglang.srt.model_executor.pool_configurator import DSV4PoolConfigurator from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _request(req_pool_idx=None, *, reused=False): diff --git a/test/registered/unit/mem_cache/test_dsv4_compress_write_pad.py b/test/registered/unit/mem_cache/test_dsv4_compress_write_pad.py index 93c2c5c0a7..272352cfe4 100644 --- a/test/registered/unit/mem_cache/test_dsv4_compress_write_pad.py +++ b/test/registered/unit/mem_cache/test_dsv4_compress_write_pad.py @@ -7,7 +7,7 @@ from sglang.srt.mem_cache.deepseek_v4_memory_pool import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestCompressStateWritePad(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_dsv4_compressed_pools.py b/test/registered/unit/mem_cache/test_dsv4_compressed_pools.py index 06e677a287..6325dbdd02 100644 --- a/test/registered/unit/mem_cache/test_dsv4_compressed_pools.py +++ b/test/registered/unit/mem_cache/test_dsv4_compressed_pools.py @@ -20,7 +20,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestDSV4CompressedPools(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_dsv4_hicache_l2.py b/test/registered/unit/mem_cache/test_dsv4_hicache_l2.py index b151a166a0..4df4a24e67 100644 --- a/test/registered/unit/mem_cache/test_dsv4_hicache_l2.py +++ b/test/registered/unit/mem_cache/test_dsv4_hicache_l2.py @@ -25,7 +25,7 @@ from sglang.srt.mem_cache.unified_cache.components import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDSV4DecodeCapacity(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_dsv4_unified_fp8_pool.py b/test/registered/unit/mem_cache/test_dsv4_unified_fp8_pool.py index 7e72aa515a..5563798583 100644 --- a/test/registered/unit/mem_cache/test_dsv4_unified_fp8_pool.py +++ b/test/registered/unit/mem_cache/test_dsv4_unified_fp8_pool.py @@ -18,7 +18,7 @@ from sglang.srt.speculative.spec_info import SpeculativeAlgorithm from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # DeepSeek-V4-Pro geometry. NOPE_DIM = 448 diff --git a/test/registered/unit/mem_cache/test_embedding_cache_controller.py b/test/registered/unit/mem_cache/test_embedding_cache_controller.py index caf170c30b..8ee1cc9814 100644 --- a/test/registered/unit/mem_cache/test_embedding_cache_controller.py +++ b/test/registered/unit/mem_cache/test_embedding_cache_controller.py @@ -21,7 +21,7 @@ from sglang.srt.mem_cache.embedding_cache_controller import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _make_pool(num_pages=16, dim=4, page_size=2, modality="vision"): diff --git a/test/registered/unit/mem_cache/test_evict_policy.py b/test/registered/unit/mem_cache/test_evict_policy.py index 1d1fbe069d..2791474bc4 100644 --- a/test/registered/unit/mem_cache/test_evict_policy.py +++ b/test/registered/unit/mem_cache/test_evict_policy.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") import unittest diff --git a/test/registered/unit/mem_cache/test_flashkda_strided_state_access.py b/test/registered/unit/mem_cache/test_flashkda_strided_state_access.py index 8d576b4f53..fe9ed76a57 100644 --- a/test/registered/unit/mem_cache/test_flashkda_strided_state_access.py +++ b/test/registered/unit/mem_cache/test_flashkda_strided_state_access.py @@ -25,7 +25,7 @@ pattern (the code under test) executes. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") import sys import types diff --git a/test/registered/unit/mem_cache/test_free_kv_row_coalesce.py b/test/registered/unit/mem_cache/test_free_kv_row_coalesce.py index 74e9e2e0fd..c35dedb143 100644 --- a/test/registered/unit/mem_cache/test_free_kv_row_coalesce.py +++ b/test/registered/unit/mem_cache/test_free_kv_row_coalesce.py @@ -17,7 +17,7 @@ from sglang.srt.mem_cache.allocator.paged import PagedTokenToKVPoolAllocator from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # A 64-token page widened by dcp_size 8 is 512; scaled down 64x: page 8, and # a Mamba track boundary at 576 tokens lands at position 9. diff --git a/test/registered/unit/mem_cache/test_full_loc_fast_path.py b/test/registered/unit/mem_cache/test_full_loc_fast_path.py index 8c301633e5..9be749f367 100644 --- a/test/registered/unit/mem_cache/test_full_loc_fast_path.py +++ b/test/registered/unit/mem_cache/test_full_loc_fast_path.py @@ -27,7 +27,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _loc_info(virtual_loc, swa_phys=None, full_phys=None): diff --git a/test/registered/unit/mem_cache/test_hicache_auto_size.py b/test/registered/unit/mem_cache/test_hicache_auto_size.py index f0ed470d52..b26be9f811 100644 --- a/test/registered/unit/mem_cache/test_hicache_auto_size.py +++ b/test/registered/unit/mem_cache/test_hicache_auto_size.py @@ -16,7 +16,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestHiCacheAutoSize(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_hicache_copy_rounds.py b/test/registered/unit/mem_cache/test_hicache_copy_rounds.py index 4893ddc058..10ce6e5c99 100644 --- a/test/registered/unit/mem_cache/test_hicache_copy_rounds.py +++ b/test/registered/unit/mem_cache/test_hicache_copy_rounds.py @@ -8,7 +8,7 @@ from unittest import mock from sglang.kernels.ops.kvcache import hicache from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _screen(element_size: int, unroll: int, *, is_hip: bool) -> bool: diff --git a/test/registered/unit/mem_cache/test_hicache_dcp_host_pool.py b/test/registered/unit/mem_cache/test_hicache_dcp_host_pool.py index af0d8db9e7..809bdb7a56 100644 --- a/test/registered/unit/mem_cache/test_hicache_dcp_host_pool.py +++ b/test/registered/unit/mem_cache/test_hicache_dcp_host_pool.py @@ -21,7 +21,7 @@ from sglang.srt.mem_cache.pool_host.mla import MLATokenToKVPoolHost from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") DCP_SIZE = 8 PHYSICAL_PAGE = 64 diff --git a/test/registered/unit/mem_cache/test_hicache_file_lru_unit.py b/test/registered/unit/mem_cache/test_hicache_file_lru_unit.py index 930741125b..a083ae3ac0 100644 --- a/test/registered/unit/mem_cache/test_hicache_file_lru_unit.py +++ b/test/registered/unit/mem_cache/test_hicache_file_lru_unit.py @@ -14,7 +14,7 @@ Run with: from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") import os import shutil diff --git a/test/registered/unit/mem_cache/test_hicache_host_register.py b/test/registered/unit/mem_cache/test_hicache_host_register.py index e12f30cd19..fe40e0e1f4 100644 --- a/test/registered/unit/mem_cache/test_hicache_host_register.py +++ b/test/registered/unit/mem_cache/test_hicache_host_register.py @@ -27,7 +27,7 @@ from sglang.srt.mem_cache.pool_host.mha import ( from sglang.srt.mem_cache.pool_host.mla import MLATokenToKVPoolHost from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeBuffer: diff --git a/test/registered/unit/mem_cache/test_hicache_load_back_timing.py b/test/registered/unit/mem_cache/test_hicache_load_back_timing.py index dbcbadf706..56bca2a35a 100644 --- a/test/registered/unit/mem_cache/test_hicache_load_back_timing.py +++ b/test/registered/unit/mem_cache/test_hicache_load_back_timing.py @@ -9,7 +9,7 @@ import torch from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small") @unittest.skipUnless(torch.cuda.is_available(), "CUDA required") diff --git a/test/registered/unit/mem_cache/test_hicache_nixl_cleaner.py b/test/registered/unit/mem_cache/test_hicache_nixl_cleaner.py index 90f6e17af6..b2dca3250d 100644 --- a/test/registered/unit/mem_cache/test_hicache_nixl_cleaner.py +++ b/test/registered/unit/mem_cache/test_hicache_nixl_cleaner.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import os import shutil diff --git a/test/registered/unit/mem_cache/test_hicache_nixl_storage.py b/test/registered/unit/mem_cache/test_hicache_nixl_storage.py index 827525493d..90ac21e66a 100644 --- a/test/registered/unit/mem_cache/test_hicache_nixl_storage.py +++ b/test/registered/unit/mem_cache/test_hicache_nixl_storage.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") import os import shutil diff --git a/test/registered/unit/mem_cache/test_hicache_staged_write_back_dispatch.py b/test/registered/unit/mem_cache/test_hicache_staged_write_back_dispatch.py index df1c6c74ec..87b76b3226 100644 --- a/test/registered/unit/mem_cache/test_hicache_staged_write_back_dispatch.py +++ b/test/registered/unit/mem_cache/test_hicache_staged_write_back_dispatch.py @@ -34,7 +34,7 @@ from sglang.srt.mem_cache.pool_host.mla import MLATokenToKVPoolHost from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") MEMORY_POOL_HOST_MODULE = "sglang.srt.mem_cache.memory_pool_host" DSA_POOL_HOST_MODULE = "sglang.srt.mem_cache.pool_host.dsa" diff --git a/test/registered/unit/mem_cache/test_hisparse_allocator.py b/test/registered/unit/mem_cache/test_hisparse_allocator.py index 837adf53cb..0781e94892 100644 --- a/test/registered/unit/mem_cache/test_hisparse_allocator.py +++ b/test/registered/unit/mem_cache/test_hisparse_allocator.py @@ -15,7 +15,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHiSparseDecodeRemap(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_hisparse_max_token_pool_size.py b/test/registered/unit/mem_cache/test_hisparse_max_token_pool_size.py index 17bd843be6..9e1d2b8fb2 100644 --- a/test/registered/unit/mem_cache/test_hisparse_max_token_pool_size.py +++ b/test/registered/unit/mem_cache/test_hisparse_max_token_pool_size.py @@ -23,7 +23,7 @@ from sglang.test.separate_buffer_allocator_double import ( ) from sglang.test.test_utils import CustomTestCase, enter_override -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_model_runner(**attrs): diff --git a/test/registered/unit/mem_cache/test_hisparse_slot_translation.py b/test/registered/unit/mem_cache/test_hisparse_slot_translation.py index b0833784d7..36fc597079 100644 --- a/test/registered/unit/mem_cache/test_hisparse_slot_translation.py +++ b/test/registered/unit/mem_cache/test_hisparse_slot_translation.py @@ -10,7 +10,7 @@ from sglang.srt.mem_cache.hisparse_memory_pool import HiSparseDSATokenToKVPool from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, stage="base-a", runner_config="cpu") +register_cpu_ci(est_time=2, stage="base-a", runner_config="cpu") class TestHiSparseSlotTranslation(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_host_memory.py b/test/registered/unit/mem_cache/test_host_memory.py index 7e8a9034f2..6c28c7f56b 100644 --- a/test/registered/unit/mem_cache/test_host_memory.py +++ b/test/registered/unit/mem_cache/test_host_memory.py @@ -8,7 +8,7 @@ from unittest.mock import Mock, patch from sglang.srt.mem_cache import host_memory from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHostMemory(unittest.TestCase): diff --git a/test/registered/unit/mem_cache/test_hybrid_pool_assembler.py b/test/registered/unit/mem_cache/test_hybrid_pool_assembler.py index 4181e6f27f..2d2a1375dc 100644 --- a/test/registered/unit/mem_cache/test_hybrid_pool_assembler.py +++ b/test/registered/unit/mem_cache/test_hybrid_pool_assembler.py @@ -21,7 +21,7 @@ from sglang.srt.mem_cache.memory_pool import HybridLinearKVPool from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDeepSeekV4SWAPageLayout(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_inkling_sconv_strided_conv_state.py b/test/registered/unit/mem_cache/test_inkling_sconv_strided_conv_state.py index 30d4e9c7eb..88ff13cc1b 100644 --- a/test/registered/unit/mem_cache/test_inkling_sconv_strided_conv_state.py +++ b/test/registered/unit/mem_cache/test_inkling_sconv_strided_conv_state.py @@ -53,7 +53,7 @@ import torch import sglang.kernels.jit as _jit_pkg from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") _KERNEL_DIR = Path(_jit_pkg.__file__).parent / "csrc" / "inkling" diff --git a/test/registered/unit/mem_cache/test_kda_fused_decode_strided_state.py b/test/registered/unit/mem_cache/test_kda_fused_decode_strided_state.py index 86a48f4697..c3428dcff1 100644 --- a/test/registered/unit/mem_cache/test_kda_fused_decode_strided_state.py +++ b/test/registered/unit/mem_cache/test_kda_fused_decode_strided_state.py @@ -33,7 +33,7 @@ arithmetic execute — no CUDA kernel is launched. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/mem_cache/test_linker_pool_assembler.py b/test/registered/unit/mem_cache/test_linker_pool_assembler.py index de912c7a17..09c1e59108 100644 --- a/test/registered/unit/mem_cache/test_linker_pool_assembler.py +++ b/test/registered/unit/mem_cache/test_linker_pool_assembler.py @@ -21,7 +21,7 @@ from sglang.srt.mem_cache.unified_cache.component_type import ComponentType from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDevicePoolEntry(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_mamba_donated_alloc_ratio.py b/test/registered/unit/mem_cache/test_mamba_donated_alloc_ratio.py index b628d176b5..9501983803 100644 --- a/test/registered/unit/mem_cache/test_mamba_donated_alloc_ratio.py +++ b/test/registered/unit/mem_cache/test_mamba_donated_alloc_ratio.py @@ -26,7 +26,7 @@ from sglang.srt.mem_cache.unified_cache.unified_tree_core import UnifiedTreeCore from sglang.srt.mem_cache.unified_radix_cache import UnifiedTreeNode from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") N = 4 # concurrent distinct-prefix requests diff --git a/test/registered/unit/mem_cache/test_mamba_path_state_cap.py b/test/registered/unit/mem_cache/test_mamba_path_state_cap.py index c22b20034d..5bd594a240 100644 --- a/test/registered/unit/mem_cache/test_mamba_path_state_cap.py +++ b/test/registered/unit/mem_cache/test_mamba_path_state_cap.py @@ -3,8 +3,8 @@ from sglang.srt.arg_groups.mamba_hook import handle_mamba_backend from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") import argparse import unittest diff --git a/test/registered/unit/mem_cache/test_mamba_state_transfer_buffers.py b/test/registered/unit/mem_cache/test_mamba_state_transfer_buffers.py index 4b889a735a..eb7e8885ca 100644 --- a/test/registered/unit/mem_cache/test_mamba_state_transfer_buffers.py +++ b/test/registered/unit/mem_cache/test_mamba_state_transfer_buffers.py @@ -5,7 +5,7 @@ import torch from sglang.srt.mem_cache.memory_pool import MambaPool from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") NUM_LAYERS = 2 NUM_SLOTS = 3 diff --git a/test/registered/unit/mem_cache/test_mamba_unittest.py b/test/registered/unit/mem_cache/test_mamba_unittest.py index e2d0c21e1e..55989779a9 100755 --- a/test/registered/unit/mem_cache/test_mamba_unittest.py +++ b/test/registered/unit/mem_cache/test_mamba_unittest.py @@ -24,7 +24,7 @@ from sglang.test.ci.ci_register import ( register_xpu_ci, ) -register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd") register_xpu_ci(est_time=20, suite="stage-b-test-1-gpu-xpu") diff --git a/test/registered/unit/mem_cache/test_mem_cache_utils.py b/test/registered/unit/mem_cache/test_mem_cache_utils.py index 0385a25dc3..3bd252e565 100644 --- a/test/registered/unit/mem_cache/test_mem_cache_utils.py +++ b/test/registered/unit/mem_cache/test_mem_cache_utils.py @@ -28,7 +28,7 @@ from sglang.srt.mem_cache.utils import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=15, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _legacy_get_hash_str(token_ids, prior_hash=None): diff --git a/test/registered/unit/mem_cache/test_mem_pool_host.py b/test/registered/unit/mem_cache/test_mem_pool_host.py index bcc54d0fbc..a34ef420f8 100644 --- a/test/registered/unit/mem_cache/test_mem_pool_host.py +++ b/test/registered/unit/mem_cache/test_mem_pool_host.py @@ -19,7 +19,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestHostKVCache(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_minimax_sparse_pool_host_unit.py b/test/registered/unit/mem_cache/test_minimax_sparse_pool_host_unit.py index 94c98edc23..e7d688cc46 100644 --- a/test/registered/unit/mem_cache/test_minimax_sparse_pool_host_unit.py +++ b/test/registered/unit/mem_cache/test_minimax_sparse_pool_host_unit.py @@ -21,7 +21,7 @@ from sglang.srt.mem_cache.pool_host.mha import ( from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/unit/mem_cache/test_minimax_sparse_pool_pd_unit.py b/test/registered/unit/mem_cache/test_minimax_sparse_pool_pd_unit.py index 96921199d6..55ae9c2bc8 100644 --- a/test/registered/unit/mem_cache/test_minimax_sparse_pool_pd_unit.py +++ b/test/registered/unit/mem_cache/test_minimax_sparse_pool_pd_unit.py @@ -12,7 +12,7 @@ from sglang.srt.mem_cache.pool_host.mha import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_k_only_pool( diff --git a/test/registered/unit/mem_cache/test_mla_host_dedup_primitives.py b/test/registered/unit/mem_cache/test_mla_host_dedup_primitives.py index 57c6905d86..a1cb8c6e72 100644 --- a/test/registered/unit/mem_cache/test_mla_host_dedup_primitives.py +++ b/test/registered/unit/mem_cache/test_mla_host_dedup_primitives.py @@ -14,7 +14,7 @@ from sglang.srt.mem_cache.pool_host.dsa import DSAIndexerPoolHost from sglang.srt.mem_cache.pool_host.mla import MLATokenToKVPoolHost from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") def _device_pool_stub(*, layer_num: int, **fields) -> SimpleNamespace: diff --git a/test/registered/unit/mem_cache/test_mmap_allocator.py b/test/registered/unit/mem_cache/test_mmap_allocator.py index 3090b1671a..83e82ff878 100644 --- a/test/registered/unit/mem_cache/test_mmap_allocator.py +++ b/test/registered/unit/mem_cache/test_mmap_allocator.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=19, suite="base-a-test-cpu") import sys diff --git a/test/registered/unit/mem_cache/test_mooncake_group_semantics.py b/test/registered/unit/mem_cache/test_mooncake_group_semantics.py index 0e814df539..67dadfd812 100644 --- a/test/registered/unit/mem_cache/test_mooncake_group_semantics.py +++ b/test/registered/unit/mem_cache/test_mooncake_group_semantics.py @@ -12,7 +12,7 @@ from sglang.srt.mem_cache.hicache_storage import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class ReplicateConfigWithGroupIds: diff --git a/test/registered/unit/mem_cache/test_mooncake_ssd_offload_path.py b/test/registered/unit/mem_cache/test_mooncake_ssd_offload_path.py index f7c19fdebe..66527520aa 100644 --- a/test/registered/unit/mem_cache/test_mooncake_ssd_offload_path.py +++ b/test/registered/unit/mem_cache/test_mooncake_ssd_offload_path.py @@ -8,7 +8,7 @@ from sglang.srt.mem_cache.hicache_storage import HiCacheStorageConfig from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _fake_store_class(): diff --git a/test/registered/unit/mem_cache/test_mooncake_standalone_dummy_mamba.py b/test/registered/unit/mem_cache/test_mooncake_standalone_dummy_mamba.py index cc17328f34..0102334c4a 100644 --- a/test/registered/unit/mem_cache/test_mooncake_standalone_dummy_mamba.py +++ b/test/registered/unit/mem_cache/test_mooncake_standalone_dummy_mamba.py @@ -5,7 +5,7 @@ from unittest.mock import patch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _fake_mooncake_modules(fake_store_cls): diff --git a/test/registered/unit/mem_cache/test_mooncake_store_config.py b/test/registered/unit/mem_cache/test_mooncake_store_config.py index 6b724cf82b..a8408b4868 100644 --- a/test/registered/unit/mem_cache/test_mooncake_store_config.py +++ b/test/registered/unit/mem_cache/test_mooncake_store_config.py @@ -16,7 +16,7 @@ from sglang.srt.utils.common import temp_set_env from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestMooncakeStoreConfigLocalHostname(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_mooncake_tenant_config.py b/test/registered/unit/mem_cache/test_mooncake_tenant_config.py index c938a18dec..339e93726e 100644 --- a/test/registered/unit/mem_cache/test_mooncake_tenant_config.py +++ b/test/registered/unit/mem_cache/test_mooncake_tenant_config.py @@ -12,7 +12,7 @@ from sglang.srt.mem_cache.storage.mooncake_store.mooncake_store import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") def _fake_mooncake_modules(fake_store_cls): diff --git a/test/registered/unit/mem_cache/test_multi_ended_allocator.py b/test/registered/unit/mem_cache/test_multi_ended_allocator.py index 2418d9bea3..94859f5e5c 100644 --- a/test/registered/unit/mem_cache/test_multi_ended_allocator.py +++ b/test/registered/unit/mem_cache/test_multi_ended_allocator.py @@ -20,7 +20,7 @@ suite runs on CPU apart from the cases that skip themselves without CUDA. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") import contextlib import random diff --git a/test/registered/unit/mem_cache/test_mxfp8_mha_pool_host_unit.py b/test/registered/unit/mem_cache/test_mxfp8_mha_pool_host_unit.py index f216c4f565..b3dfbda4d5 100644 --- a/test/registered/unit/mem_cache/test_mxfp8_mha_pool_host_unit.py +++ b/test/registered/unit/mem_cache/test_mxfp8_mha_pool_host_unit.py @@ -16,7 +16,7 @@ from sglang.srt.mem_cache.pool_host.mha_mxfp8 import MHATokenToKVPoolMXFP8Host from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") MXFP8_MODULE = "sglang.srt.mem_cache.pool_host.mha_mxfp8" MHA_MODULE = "sglang.srt.mem_cache.pool_host.mha" diff --git a/test/registered/unit/mem_cache/test_mxfp8_scale_transfer_buffers.py b/test/registered/unit/mem_cache/test_mxfp8_scale_transfer_buffers.py index 5484c7ae42..22a7743fb9 100644 --- a/test/registered/unit/mem_cache/test_mxfp8_scale_transfer_buffers.py +++ b/test/registered/unit/mem_cache/test_mxfp8_scale_transfer_buffers.py @@ -5,7 +5,7 @@ import torch from sglang.srt.mem_cache.memory_pool import MHATokenToKVPoolMXFP8 from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") LAYERS = 2 PAGE_SIZE = 128 diff --git a/test/registered/unit/mem_cache/test_npu_mamba_async_layout.py b/test/registered/unit/mem_cache/test_npu_mamba_async_layout.py index 01b05bd960..bc8e434fb3 100644 --- a/test/registered/unit/mem_cache/test_npu_mamba_async_layout.py +++ b/test/registered/unit/mem_cache/test_npu_mamba_async_layout.py @@ -12,7 +12,7 @@ from sglang.srt.mem_cache.pool_host.mamba import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class TestNPUMambaAsyncConfig(unittest.TestCase): diff --git a/test/registered/unit/mem_cache/test_page_interleave_shard.py b/test/registered/unit/mem_cache/test_page_interleave_shard.py index 8ddcd37d43..cab687bca3 100644 --- a/test/registered/unit/mem_cache/test_page_interleave_shard.py +++ b/test/registered/unit/mem_cache/test_page_interleave_shard.py @@ -86,7 +86,7 @@ from sglang.srt.utils import ceil_div from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase, publish_build_topology -register_cpu_ci(est_time=30, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") N = 4 # shard size PS = 16 # physical page size diff --git a/test/registered/unit/mem_cache/test_paged_allocator_lazy_release.py b/test/registered/unit/mem_cache/test_paged_allocator_lazy_release.py index 908e3ecb6a..50d2fde194 100644 --- a/test/registered/unit/mem_cache/test_paged_allocator_lazy_release.py +++ b/test/registered/unit/mem_cache/test_paged_allocator_lazy_release.py @@ -9,7 +9,7 @@ from sglang.srt.mem_cache.allocator.paged import PagedTokenToKVPoolAllocator from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") PAGE_SIZE = 2 NUM_PAGES = 16 diff --git a/test/registered/unit/mem_cache/test_paged_free_segment.py b/test/registered/unit/mem_cache/test_paged_free_segment.py index 30bdbac062..61678d5dd6 100644 --- a/test/registered/unit/mem_cache/test_paged_free_segment.py +++ b/test/registered/unit/mem_cache/test_paged_free_segment.py @@ -17,7 +17,7 @@ from sglang.srt.mem_cache.allocator.paged import PagedTokenToKVPoolAllocator from sglang.srt.mem_cache.common import _release_overallocated_kv_indices from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") PAGE_SIZE = 4 NUM_PAGES = 64 diff --git a/test/registered/unit/mem_cache/test_pd_envelope_transfer_layout.py b/test/registered/unit/mem_cache/test_pd_envelope_transfer_layout.py index 8ecfb96767..fa769e1109 100644 --- a/test/registered/unit/mem_cache/test_pd_envelope_transfer_layout.py +++ b/test/registered/unit/mem_cache/test_pd_envelope_transfer_layout.py @@ -28,7 +28,7 @@ from sglang.srt.mem_cache.layout.page_major import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestMLAEnvelopeTransferAddressing(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_pp_prefetch_ticket.py b/test/registered/unit/mem_cache/test_pp_prefetch_ticket.py index 6802af04e6..d3a51ff99a 100644 --- a/test/registered/unit/mem_cache/test_pp_prefetch_ticket.py +++ b/test/registered/unit/mem_cache/test_pp_prefetch_ticket.py @@ -26,7 +26,7 @@ from sglang.srt.mem_cache.unified_radix_cache import UnifiedRadixCache from sglang.srt.mem_cache.utils import get_storage_hash_str from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") class TestPPPrefetchTicket(unittest.TestCase): diff --git a/test/registered/unit/mem_cache/test_prefill_memory_budget.py b/test/registered/unit/mem_cache/test_prefill_memory_budget.py index 2a73dfffc9..825baa66cc 100644 --- a/test/registered/unit/mem_cache/test_prefill_memory_budget.py +++ b/test/registered/unit/mem_cache/test_prefill_memory_budget.py @@ -27,7 +27,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _shared_allocator(*, page_size=4, total_bytes=1024): diff --git a/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py b/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py index 25d3911cf3..5ebf013dd4 100644 --- a/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py +++ b/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py @@ -9,7 +9,7 @@ from sglang.srt.mem_cache.pure_swa_radix_cache import PureSWARadixCache from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeAllocator: diff --git a/test/registered/unit/mem_cache/test_quantized_kv_pool.py b/test/registered/unit/mem_cache/test_quantized_kv_pool.py index 51da4b8964..8885443e48 100644 --- a/test/registered/unit/mem_cache/test_quantized_kv_pool.py +++ b/test/registered/unit/mem_cache/test_quantized_kv_pool.py @@ -8,7 +8,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") class _FakeQuantMethod: diff --git a/test/registered/unit/mem_cache/test_radix_cache_unit.py b/test/registered/unit/mem_cache/test_radix_cache_unit.py index 31f165402a..3081bbb5c3 100644 --- a/test/registered/unit/mem_cache/test_radix_cache_unit.py +++ b/test/registered/unit/mem_cache/test_radix_cache_unit.py @@ -20,7 +20,7 @@ Usage: from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci # CPU-based unit test, runs quickly on any GPU runner -register_cuda_ci(est_time=14, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=18, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=5, suite="stage-b-test-1-gpu-small-amd") import random diff --git a/test/registered/unit/mem_cache/test_radix_force_miss.py b/test/registered/unit/mem_cache/test_radix_force_miss.py index 38d670712e..34e79109d1 100644 --- a/test/registered/unit/mem_cache/test_radix_force_miss.py +++ b/test/registered/unit/mem_cache/test_radix_force_miss.py @@ -7,7 +7,7 @@ RadixCache. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") import unittest import unittest.mock diff --git a/test/registered/unit/mem_cache/test_registry.py b/test/registered/unit/mem_cache/test_registry.py index 47bb5671cf..99917003ed 100644 --- a/test/registered/unit/mem_cache/test_registry.py +++ b/test/registered/unit/mem_cache/test_registry.py @@ -3,7 +3,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") import unittest from unittest.mock import MagicMock, patch diff --git a/test/registered/unit/mem_cache/test_replayssm_ring_accounting.py b/test/registered/unit/mem_cache/test_replayssm_ring_accounting.py index 6377f6c112..de3644b391 100644 --- a/test/registered/unit/mem_cache/test_replayssm_ring_accounting.py +++ b/test/registered/unit/mem_cache/test_replayssm_ring_accounting.py @@ -23,7 +23,7 @@ from sglang.srt.mem_cache.kv_cache_configurator import _pp_local_per_request_byt from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") # temporal = (hv=4, v_dim=8, k_dim=8), num_k_heads_per_tp = 4, record_len = 8, # 2 layers. conv bf16 (2B), fp32 gate/beta (4B). Ring tensors (per slot, per diff --git a/test/registered/unit/mem_cache/test_retraction_mamba_backup.py b/test/registered/unit/mem_cache/test_retraction_mamba_backup.py index adc138b6de..0ba6522e66 100644 --- a/test/registered/unit/mem_cache/test_retraction_mamba_backup.py +++ b/test/registered/unit/mem_cache/test_retraction_mamba_backup.py @@ -6,7 +6,7 @@ from sglang.srt.managers.schedule_batch import Req, ReqKvInfo from sglang.srt.mem_cache.memory_pool import HybridReqToTokenPool from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") MAMBA_STATE = object() diff --git a/test/registered/unit/mem_cache/test_rust_tree_core_integration.py b/test/registered/unit/mem_cache/test_rust_tree_core_integration.py index 6e710ed370..f31ba7ab77 100644 --- a/test/registered/unit/mem_cache/test_rust_tree_core_integration.py +++ b/test/registered/unit/mem_cache/test_rust_tree_core_integration.py @@ -11,7 +11,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=17, suite="base-a-test-cpu") +register_cpu_ci(est_time=14, suite="base-a-test-cpu") if shutil.which("cargo") is None: pytest.skip("the rust backend builds with cargo", allow_module_level=True) diff --git a/test/registered/unit/mem_cache/test_rust_unified_radix_cache_bench.py b/test/registered/unit/mem_cache/test_rust_unified_radix_cache_bench.py index 9fbd2dae5b..af1199c957 100644 --- a/test/registered/unit/mem_cache/test_rust_unified_radix_cache_bench.py +++ b/test/registered/unit/mem_cache/test_rust_unified_radix_cache_bench.py @@ -6,7 +6,7 @@ import test_unified_radix_cache_bench as shared_suite from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=25, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=27, stage="base-b", runner_config="1-gpu-small") class RustBackendSuite(unittest.TestSuite): diff --git a/test/registered/unit/mem_cache/test_session_token_share_unit.py b/test/registered/unit/mem_cache/test_session_token_share_unit.py index cf12e063c6..7bddb851d3 100644 --- a/test/registered/unit/mem_cache/test_session_token_share_unit.py +++ b/test/registered/unit/mem_cache/test_session_token_share_unit.py @@ -10,7 +10,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest from array import array diff --git a/test/registered/unit/mem_cache/test_session_unified_radix_cache.py b/test/registered/unit/mem_cache/test_session_unified_radix_cache.py index e50475985d..41a7d8fd1f 100644 --- a/test/registered/unit/mem_cache/test_session_unified_radix_cache.py +++ b/test/registered/unit/mem_cache/test_session_unified_radix_cache.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest from array import array diff --git a/test/registered/unit/mem_cache/test_storage_prefetch_lifecycle.py b/test/registered/unit/mem_cache/test_storage_prefetch_lifecycle.py index 29331c03f5..892a626660 100644 --- a/test/registered/unit/mem_cache/test_storage_prefetch_lifecycle.py +++ b/test/registered/unit/mem_cache/test_storage_prefetch_lifecycle.py @@ -32,7 +32,7 @@ from sglang.srt.mem_cache.unified_radix_cache import ( from sglang.srt.mem_cache.utils import get_hash_str from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=20, suite="base-a-test-cpu") +register_cpu_ci(est_time=84, suite="base-a-test-cpu") _REQ = CacheRequestHandle("r", 0) diff --git a/test/registered/unit/mem_cache/test_streaming_session_unit.py b/test/registered/unit/mem_cache/test_streaming_session_unit.py index 166106dd3e..2fa984f6e7 100644 --- a/test/registered/unit/mem_cache/test_streaming_session_unit.py +++ b/test/registered/unit/mem_cache/test_streaming_session_unit.py @@ -8,7 +8,7 @@ from sglang.srt.mem_cache.base_prefix_cache import DecLockRefParams, MatchResult from sglang.srt.session.streaming_session import SessionSlot, StreamingSession from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeAllocator(BaseTokenToKVPoolAllocator): diff --git a/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py b/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py index 08c153e357..641198e321 100644 --- a/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py +++ b/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py @@ -15,7 +15,7 @@ from sglang.srt.mem_cache.allocator.swa import SWATokenToKVPoolAllocator from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _make_self(*, page_size: int, full_available: int, swa_available: int): diff --git a/test/registered/unit/mem_cache/test_swa_locked_full_recover_unified.py b/test/registered/unit/mem_cache/test_swa_locked_full_recover_unified.py index 2be88d3516..df7dc0a17d 100644 --- a/test/registered/unit/mem_cache/test_swa_locked_full_recover_unified.py +++ b/test/registered/unit/mem_cache/test_swa_locked_full_recover_unified.py @@ -41,7 +41,7 @@ from sglang.srt.mem_cache.unified_cache.components.swa import SWAComponent from sglang.srt.mem_cache.unified_memory_pool import MHASubPoolSpec, UnifiedKVPool from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") _DEV = "cpu" _SWA = ComponentType.SWA diff --git a/test/registered/unit/mem_cache/test_swa_pool_v_head_dim.py b/test/registered/unit/mem_cache/test_swa_pool_v_head_dim.py index d3fd443af4..7a3b244716 100644 --- a/test/registered/unit/mem_cache/test_swa_pool_v_head_dim.py +++ b/test/registered/unit/mem_cache/test_swa_pool_v_head_dim.py @@ -27,7 +27,7 @@ import torch from sglang.srt.mem_cache.swa_memory_pool import SWAKVPool from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") _DEV = "cpu" _FULL_V_HEAD_DIM = 8 diff --git a/test/registered/unit/mem_cache/test_swa_ring_page_return.py b/test/registered/unit/mem_cache/test_swa_ring_page_return.py index 858a49d86f..fad7062e28 100644 --- a/test/registered/unit/mem_cache/test_swa_ring_page_return.py +++ b/test/registered/unit/mem_cache/test_swa_ring_page_return.py @@ -19,7 +19,7 @@ from sglang.srt.mem_cache.allocator.swa import SWATokenToKVPoolAllocator from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") PAGE_SIZE = 8 POOL_PAGES = 4 diff --git a/test/registered/unit/mem_cache/test_tlru_eviction_policy.py b/test/registered/unit/mem_cache/test_tlru_eviction_policy.py index 0e56fdf241..a1f97816d2 100644 --- a/test/registered/unit/mem_cache/test_tlru_eviction_policy.py +++ b/test/registered/unit/mem_cache/test_tlru_eviction_policy.py @@ -24,7 +24,7 @@ except ImportError: # standalone run without an sglang install; CI parses the pass -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") _HERE = os.path.dirname(os.path.abspath(__file__)) _EVICT_POLICY = os.path.normpath( diff --git a/test/registered/unit/mem_cache/test_tree_core_registry.py b/test/registered/unit/mem_cache/test_tree_core_registry.py index 11c8b9d591..237b18f1d7 100644 --- a/test/registered/unit/mem_cache/test_tree_core_registry.py +++ b/test/registered/unit/mem_cache/test_tree_core_registry.py @@ -24,7 +24,7 @@ from sglang.srt.mem_cache.unified_radix_cache import UnifiedRadixCache from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _cache_init_params(**kwargs) -> CacheInitParams: diff --git a/test/registered/unit/mem_cache/test_umbp_host_allocator.py b/test/registered/unit/mem_cache/test_umbp_host_allocator.py index 2c02f29579..fcc16f6066 100644 --- a/test/registered/unit/mem_cache/test_umbp_host_allocator.py +++ b/test/registered/unit/mem_cache/test_umbp_host_allocator.py @@ -12,7 +12,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # These tests stub out mori with a fake in-process module, so they need neither # a real mori install nor a GPU and run on NVIDIA / CPU CI. diff --git a/test/registered/unit/mem_cache/test_unified_byte_accounting.py b/test/registered/unit/mem_cache/test_unified_byte_accounting.py index ee7a7e74a7..4c6f70b702 100644 --- a/test/registered/unit/mem_cache/test_unified_byte_accounting.py +++ b/test/registered/unit/mem_cache/test_unified_byte_accounting.py @@ -36,7 +36,7 @@ from sglang.srt.runtime_context import publish, reset_context from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") def _swa_composite(): diff --git a/test/registered/unit/mem_cache/test_unified_cache_linker.py b/test/registered/unit/mem_cache/test_unified_cache_linker.py index fe6ba8ce9b..1c435aeca7 100644 --- a/test/registered/unit/mem_cache/test_unified_cache_linker.py +++ b/test/registered/unit/mem_cache/test_unified_cache_linker.py @@ -50,8 +50,8 @@ from sglang.srt.mem_cache.unified_cache.unified_cache_linker import ( from sglang.srt.mem_cache.unified_radix_cache import UnifiedRadixCache from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") -register_cuda_ci(est_time=100, stage="base-b", runner_config="1-gpu-small") +register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cuda_ci(est_time=19, stage="base-b", runner_config="1-gpu-small") class _FakeLinker(UnifiedCacheLinker): diff --git a/test/registered/unit/mem_cache/test_unified_capacity_memo.py b/test/registered/unit/mem_cache/test_unified_capacity_memo.py index d28939aead..eb524633e5 100644 --- a/test/registered/unit/mem_cache/test_unified_capacity_memo.py +++ b/test/registered/unit/mem_cache/test_unified_capacity_memo.py @@ -29,7 +29,7 @@ import unittest from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _build(lazy: bool): diff --git a/test/registered/unit/mem_cache/test_unified_free_no_host_sync.py b/test/registered/unit/mem_cache/test_unified_free_no_host_sync.py index 05063d7107..408d7bad60 100644 --- a/test/registered/unit/mem_cache/test_unified_free_no_host_sync.py +++ b/test/registered/unit/mem_cache/test_unified_free_no_host_sync.py @@ -38,7 +38,7 @@ from sglang.srt.mem_cache.allocator import unified_sub_pool as mea from sglang.srt.mem_cache.allocator.base import BaseTokenToKVPoolAllocator from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") PAGE_SIZE = _PagedFixture.PAGE_SIZE diff --git a/test/registered/unit/mem_cache/test_unified_handout_zeroing.py b/test/registered/unit/mem_cache/test_unified_handout_zeroing.py index 51d46c38e8..40decfd531 100644 --- a/test/registered/unit/mem_cache/test_unified_handout_zeroing.py +++ b/test/registered/unit/mem_cache/test_unified_handout_zeroing.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-small") import unittest diff --git a/test/registered/unit/mem_cache/test_unified_hicache_regressions.py b/test/registered/unit/mem_cache/test_unified_hicache_regressions.py index 4bd2da3a9d..6557f38dd5 100644 --- a/test/registered/unit/mem_cache/test_unified_hicache_regressions.py +++ b/test/registered/unit/mem_cache/test_unified_hicache_regressions.py @@ -18,7 +18,7 @@ from sglang.srt.mem_cache.unified_cache.component_type import ComponentType from sglang.srt.mem_cache.unified_cache.unified_tree_core import UnifiedTreeCore from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=15, stage="extra-a", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="extra-a", runner_config="1-gpu-small") class TestHiCacheIndexDomains(unittest.TestCase): diff --git a/test/registered/unit/mem_cache/test_unified_mamba_views.py b/test/registered/unit/mem_cache/test_unified_mamba_views.py index b9f1eef0a3..b02c015353 100644 --- a/test/registered/unit/mem_cache/test_unified_mamba_views.py +++ b/test/registered/unit/mem_cache/test_unified_mamba_views.py @@ -31,7 +31,7 @@ import torch from sglang.test.ci.ci_register import register_amd_ci, register_cpu_ci -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_amd_ci(est_time=30, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/unit/mem_cache/test_unified_mha_views.py b/test/registered/unit/mem_cache/test_unified_mha_views.py index 1c400900d9..666b8494bc 100644 --- a/test/registered/unit/mem_cache/test_unified_mha_views.py +++ b/test/registered/unit/mem_cache/test_unified_mha_views.py @@ -23,7 +23,7 @@ Addressing law under test: from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/mem_cache/test_unified_mla_block_table.py b/test/registered/unit/mem_cache/test_unified_mla_block_table.py index a8204c7e31..6fe54601b5 100644 --- a/test/registered/unit/mem_cache/test_unified_mla_block_table.py +++ b/test/registered/unit/mem_cache/test_unified_mla_block_table.py @@ -53,7 +53,7 @@ import torch from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=8, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=6, stage="base-b", runner_config="1-gpu-small") _HAS_CUDA = torch.cuda.is_available() _DEV = "cuda" diff --git a/test/registered/unit/mem_cache/test_unified_mla_gpu_parity.py b/test/registered/unit/mem_cache/test_unified_mla_gpu_parity.py index 9cb7996903..b868a42608 100644 --- a/test/registered/unit/mem_cache/test_unified_mla_gpu_parity.py +++ b/test/registered/unit/mem_cache/test_unified_mla_gpu_parity.py @@ -30,7 +30,7 @@ from sglang.srt.runtime_context import publish, reset_context from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") _HAS_CUDA = torch.cuda.is_available() _DEV = "cuda" diff --git a/test/registered/unit/mem_cache/test_unified_mla_views.py b/test/registered/unit/mem_cache/test_unified_mla_views.py index 0c0347327d..5c37aca049 100644 --- a/test/registered/unit/mem_cache/test_unified_mla_views.py +++ b/test/registered/unit/mem_cache/test_unified_mla_views.py @@ -23,7 +23,7 @@ GPU parity of the read/write kernels (set_mla_kv_buffer TMA path etc.) lives in from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from unittest import mock diff --git a/test/registered/unit/mem_cache/test_unified_npool_sweep.py b/test/registered/unit/mem_cache/test_unified_npool_sweep.py index 6af10bdd44..fb0b18a93b 100644 --- a/test/registered/unit/mem_cache/test_unified_npool_sweep.py +++ b/test/registered/unit/mem_cache/test_unified_npool_sweep.py @@ -36,7 +36,7 @@ from sglang.test.ci.ci_register import register_cpu_ci # Hermetic convention of this directory's pool tests: plain unittest.TestCase, # only ci_register imported (no heavy sglang.test.test_utils chain). -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") _DEV = "cpu" diff --git a/test/registered/unit/mem_cache/test_unified_radix_allocation_eviction.py b/test/registered/unit/mem_cache/test_unified_radix_allocation_eviction.py index 5b99273aa8..fcec307a0c 100644 --- a/test/registered/unit/mem_cache/test_unified_radix_allocation_eviction.py +++ b/test/registered/unit/mem_cache/test_unified_radix_allocation_eviction.py @@ -18,7 +18,7 @@ from sglang.srt.mem_cache.unified_radix_cache import UnifiedRadixCache from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestUnifiedRadixAllocationEviction(CustomTestCase): diff --git a/test/registered/unit/mem_cache/test_unified_radix_hicache_dispatch.py b/test/registered/unit/mem_cache/test_unified_radix_hicache_dispatch.py index f132348175..54f91dc8fa 100644 --- a/test/registered/unit/mem_cache/test_unified_radix_hicache_dispatch.py +++ b/test/registered/unit/mem_cache/test_unified_radix_hicache_dispatch.py @@ -20,7 +20,7 @@ from sglang.srt.mem_cache.hybrid_cache.hybrid_pool_assembler import ( from sglang.srt.mem_cache.unified_cache.components import ComponentType from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _mock_kvcache(cls): diff --git a/test/registered/unit/mem_cache/test_unified_tri_pool.py b/test/registered/unit/mem_cache/test_unified_tri_pool.py index db1f31b066..0e83ea7b61 100644 --- a/test/registered/unit/mem_cache/test_unified_tri_pool.py +++ b/test/registered/unit/mem_cache/test_unified_tri_pool.py @@ -45,7 +45,7 @@ from sglang.test.ci.ci_register import register_cpu_ci # Hermetic convention of this directory's pool tests: plain unittest.TestCase, # only ci_register imported (no heavy sglang.test.test_utils chain). -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _DEV = "cpu" diff --git a/test/registered/unit/memory/test_ptr_table.py b/test/registered/unit/memory/test_ptr_table.py index 0f5d090613..97ca0dc0a0 100644 --- a/test/registered/unit/memory/test_ptr_table.py +++ b/test/registered/unit/memory/test_ptr_table.py @@ -7,7 +7,7 @@ real device memory lives in test/registered/xpu/test_ptr_table.py. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/model_executor/model_runner_components/test_attention_backend_setup.py b/test/registered/unit/model_executor/model_runner_components/test_attention_backend_setup.py index 426145ea36..22a62c0c4f 100644 --- a/test/registered/unit/model_executor/model_runner_components/test_attention_backend_setup.py +++ b/test/registered/unit/model_executor/model_runner_components/test_attention_backend_setup.py @@ -13,7 +13,7 @@ from sglang.srt.model_executor.model_runner_components.attention_backend_setup i ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeBackend: diff --git a/test/registered/unit/model_executor/model_runner_components/test_cuda_graph_setup.py b/test/registered/unit/model_executor/model_runner_components/test_cuda_graph_setup.py index ac9aad7da8..a5f413b0e8 100644 --- a/test/registered/unit/model_executor/model_runner_components/test_cuda_graph_setup.py +++ b/test/registered/unit/model_executor/model_runner_components/test_cuda_graph_setup.py @@ -12,7 +12,7 @@ from sglang.srt.model_executor.model_runner_components.cuda_graph_setup import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def test_standard_gqa_gate_uses_pipeline_local_layer_range(): diff --git a/test/registered/unit/model_executor/model_runner_components/test_layer_setup.py b/test/registered/unit/model_executor/model_runner_components/test_layer_setup.py index 94faa44d6d..80cf30fea6 100644 --- a/test/registered/unit/model_executor/model_runner_components/test_layer_setup.py +++ b/test/registered/unit/model_executor/model_runner_components/test_layer_setup.py @@ -8,7 +8,7 @@ from sglang.srt.model_executor.model_runner_components.layer_setup import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestComputeAttentionAndMoeLayers(unittest.TestCase): diff --git a/test/registered/unit/model_executor/model_runner_components/test_ngram_embedding_manager.py b/test/registered/unit/model_executor/model_runner_components/test_ngram_embedding_manager.py index 477c67930c..4798275108 100644 --- a/test/registered/unit/model_executor/model_runner_components/test_ngram_embedding_manager.py +++ b/test/registered/unit/model_executor/model_runner_components/test_ngram_embedding_manager.py @@ -15,7 +15,7 @@ from sglang.srt.model_executor.model_runner_components.ngram_embedding_manager i update_ngram_token_table_after_sampling, ) -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_ngram_info(batch_size: int, skip_token_table_update=None): diff --git a/test/registered/unit/model_executor/model_runner_components/test_spec_aux_hidden_state.py b/test/registered/unit/model_executor/model_runner_components/test_spec_aux_hidden_state.py index a48604188f..252b410cf8 100644 --- a/test/registered/unit/model_executor/model_runner_components/test_spec_aux_hidden_state.py +++ b/test/registered/unit/model_executor/model_runner_components/test_spec_aux_hidden_state.py @@ -8,7 +8,7 @@ from sglang.srt.model_executor.model_runner_components.spec_aux_hidden_state imp ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") @pytest.mark.parametrize( diff --git a/test/registered/unit/model_executor/model_runner_components/test_startup_weight_load.py b/test/registered/unit/model_executor/model_runner_components/test_startup_weight_load.py index 0c3ee44f70..a0d5316da2 100644 --- a/test/registered/unit/model_executor/model_runner_components/test_startup_weight_load.py +++ b/test/registered/unit/model_executor/model_runner_components/test_startup_weight_load.py @@ -32,7 +32,7 @@ from sglang.srt.model_loader.weight_utils import initialize_capture_safe_weights from sglang.srt.runtime_context import get_context, publish, reset_context from sglang.srt.server_args import ServerArgs -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") _STARTUP_MODULE = ( diff --git a/test/registered/unit/model_executor/runner/test_decode_cuda_graph_runner.py b/test/registered/unit/model_executor/runner/test_decode_cuda_graph_runner.py index 03990cd682..fd8ee1f27c 100644 --- a/test/registered/unit/model_executor/runner/test_decode_cuda_graph_runner.py +++ b/test/registered/unit/model_executor/runner/test_decode_cuda_graph_runner.py @@ -36,7 +36,7 @@ from sglang.srt.utils import profile_utils as putils from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _CAPTURE_TRACE = "SGLANG_ENABLE_CUDA_GRAPH_CAPTURE_TRACE" _BATCH_CAPTURE = "SGLANG_GRAPH_BATCH_CAPTURE" diff --git a/test/registered/unit/model_executor/runner/test_decode_cuda_graph_shared_read_fence.py b/test/registered/unit/model_executor/runner/test_decode_cuda_graph_shared_read_fence.py index 337408a25e..4d7d69e6e7 100644 --- a/test/registered/unit/model_executor/runner/test_decode_cuda_graph_shared_read_fence.py +++ b/test/registered/unit/model_executor/runner/test_decode_cuda_graph_shared_read_fence.py @@ -13,7 +13,7 @@ from sglang.srt.model_executor.runner.decode_cuda_graph_runner import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") DECODE = ForwardMode.DECODE diff --git a/test/registered/unit/model_executor/runner/test_flashinfer_autotune_sync.py b/test/registered/unit/model_executor/runner/test_flashinfer_autotune_sync.py index 930450a831..08fa217a28 100644 --- a/test/registered/unit/model_executor/runner/test_flashinfer_autotune_sync.py +++ b/test/registered/unit/model_executor/runner/test_flashinfer_autotune_sync.py @@ -8,7 +8,7 @@ cover that gate and the digest it decides on. from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci -register_cpu_ci(est_time=57, suite="base-a-test-cpu") +register_cpu_ci(est_time=44, suite="base-a-test-cpu") register_cuda_ci(est_time=25, stage="base-b-kernel-unit", runner_config="1-gpu-large") import json diff --git a/test/registered/unit/model_executor/runner/test_hidden_state_graph_recapture.py b/test/registered/unit/model_executor/runner/test_hidden_state_graph_recapture.py index 7581b8f7a9..9d3ce14d6d 100644 --- a/test/registered/unit/model_executor/runner/test_hidden_state_graph_recapture.py +++ b/test/registered/unit/model_executor/runner/test_hidden_state_graph_recapture.py @@ -19,7 +19,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHiddenStateGraphRecapture(CustomTestCase): diff --git a/test/registered/unit/model_executor/runner/test_prefill_cuda_graph_padding.py b/test/registered/unit/model_executor/runner/test_prefill_cuda_graph_padding.py index ed9ebdeda7..9ab75be207 100644 --- a/test/registered/unit/model_executor/runner/test_prefill_cuda_graph_padding.py +++ b/test/registered/unit/model_executor/runner/test_prefill_cuda_graph_padding.py @@ -21,7 +21,7 @@ from sglang.srt.model_executor.runner.shape_key import ShapeKey from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestPrefillCudaGraphPadding(CustomTestCase): diff --git a/test/registered/unit/model_executor/runner/test_prefill_shared_read_done.py b/test/registered/unit/model_executor/runner/test_prefill_shared_read_done.py index 0c96e270aa..c675272ff3 100644 --- a/test/registered/unit/model_executor/runner/test_prefill_shared_read_done.py +++ b/test/registered/unit/model_executor/runner/test_prefill_shared_read_done.py @@ -17,7 +17,7 @@ from sglang.srt.speculative.spec_registry import CustomSpecAlgo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _Event: diff --git a/test/registered/unit/model_executor/runner_backend/test_backend_resolution.py b/test/registered/unit/model_executor/runner_backend/test_backend_resolution.py index a78725a5e6..1ca5f84ed5 100644 --- a/test/registered/unit/model_executor/runner_backend/test_backend_resolution.py +++ b/test/registered/unit/model_executor/runner_backend/test_backend_resolution.py @@ -11,7 +11,7 @@ from sglang.srt.platforms.interface import SRTPlatform from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_graph_runner(*, device="custom"): diff --git a/test/registered/unit/model_executor/runner_backend/test_full_cuda_graph_backend.py b/test/registered/unit/model_executor/runner_backend/test_full_cuda_graph_backend.py index f653c8d9b0..3ef0553ba1 100644 --- a/test/registered/unit/model_executor/runner_backend/test_full_cuda_graph_backend.py +++ b/test/registered/unit/model_executor/runner_backend/test_full_cuda_graph_backend.py @@ -33,7 +33,7 @@ from sglang.srt.model_executor.runner_backend.full_cuda_graph_backend import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # Sentinel: distinguishes "runner has no _profiler attribute" from # "_profiler is None" in the test fixtures. diff --git a/test/registered/unit/model_executor/runner_utils/test_graph_pool_borrow.py b/test/registered/unit/model_executor/runner_utils/test_graph_pool_borrow.py index 784c91b0b8..d72b1fc25b 100644 --- a/test/registered/unit/model_executor/runner_utils/test_graph_pool_borrow.py +++ b/test/registered/unit/model_executor/runner_utils/test_graph_pool_borrow.py @@ -24,7 +24,7 @@ from sglang.srt.speculative.dflash_worker_v2 import DFlashWorkerV2 from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=26, stage="base-b", runner_config="1-gpu-small") class TestGraphPoolBorrow(CustomTestCase): diff --git a/test/registered/unit/model_executor/test_chunked_prefix_cache_gate.py b/test/registered/unit/model_executor/test_chunked_prefix_cache_gate.py index c689f3d921..d777baf827 100644 --- a/test/registered/unit/model_executor/test_chunked_prefix_cache_gate.py +++ b/test/registered/unit/model_executor/test_chunked_prefix_cache_gate.py @@ -7,7 +7,7 @@ saw the flip, so an unsupported backend kept chunked prefix enabled. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/model_executor/test_compensated_mhc_update_guard.py b/test/registered/unit/model_executor/test_compensated_mhc_update_guard.py index 82e27cca44..76ad975dc2 100644 --- a/test/registered/unit/model_executor/test_compensated_mhc_update_guard.py +++ b/test/registered/unit/model_executor/test_compensated_mhc_update_guard.py @@ -14,7 +14,7 @@ from sglang.srt.model_executor.model_runner_components.weight_updater import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestCompensatedMhcUpdateGuard(CustomTestCase): diff --git a/test/registered/unit/model_executor/test_cuda_graph_buffer_registry.py b/test/registered/unit/model_executor/test_cuda_graph_buffer_registry.py index b7ef20aaa9..9eacaa777b 100644 --- a/test/registered/unit/model_executor/test_cuda_graph_buffer_registry.py +++ b/test/registered/unit/model_executor/test_cuda_graph_buffer_registry.py @@ -29,7 +29,7 @@ from sglang.srt.model_executor.cuda_graph_buffer_registry import ( from sglang.srt.model_executor.input_buffers import ForwardInputBuffers from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") @dataclasses.dataclass diff --git a/test/registered/unit/model_executor/test_forward_metadata_plan_record.py b/test/registered/unit/model_executor/test_forward_metadata_plan_record.py index ed8692127a..bb936f7638 100644 --- a/test/registered/unit/model_executor/test_forward_metadata_plan_record.py +++ b/test/registered/unit/model_executor/test_forward_metadata_plan_record.py @@ -21,7 +21,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch, ForwardMo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_batch(bs: int = 2, num_tokens: int = 2) -> ForwardBatch: diff --git a/test/registered/unit/model_executor/test_hisparse_pool_configurator.py b/test/registered/unit/model_executor/test_hisparse_pool_configurator.py index 7789072191..98456a7468 100644 --- a/test/registered/unit/model_executor/test_hisparse_pool_configurator.py +++ b/test/registered/unit/model_executor/test_hisparse_pool_configurator.py @@ -9,7 +9,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHiSparsePoolConfigurator(CustomTestCase): diff --git a/test/registered/unit/model_executor/test_kv_canary_headroom.py b/test/registered/unit/model_executor/test_kv_canary_headroom.py index 132092ad40..2377b508e4 100644 --- a/test/registered/unit/model_executor/test_kv_canary_headroom.py +++ b/test/registered/unit/model_executor/test_kv_canary_headroom.py @@ -15,7 +15,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, stage="base-a", runner_config="cpu") +register_cpu_ci(est_time=2, stage="base-a", runner_config="cpu") _GIB = 1 << 30 diff --git a/test/registered/unit/model_executor/test_mlp_sync_pad_unpad.py b/test/registered/unit/model_executor/test_mlp_sync_pad_unpad.py index a7c9565e19..0b5140260e 100644 --- a/test/registered/unit/model_executor/test_mlp_sync_pad_unpad.py +++ b/test/registered/unit/model_executor/test_mlp_sync_pad_unpad.py @@ -23,7 +23,7 @@ from sglang.srt.model_executor.runner.decode_cuda_graph_runner import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _mock_model_runner(seq_len_fill_value: int = 1) -> MagicMock: diff --git a/test/registered/unit/model_executor/test_model_runner_decode_rows.py b/test/registered/unit/model_executor/test_model_runner_decode_rows.py index fe0668ba81..17c3d767f8 100644 --- a/test/registered/unit/model_executor/test_model_runner_decode_rows.py +++ b/test/registered/unit/model_executor/test_model_runner_decode_rows.py @@ -6,7 +6,7 @@ from unittest.mock import patch from sglang.srt.model_executor.model_runner import ModelRunner from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeModelRunner: diff --git a/test/registered/unit/model_executor/test_num_token_non_padded_localization.py b/test/registered/unit/model_executor/test_num_token_non_padded_localization.py index b8446fc686..66ac509dfc 100644 --- a/test/registered/unit/model_executor/test_num_token_non_padded_localization.py +++ b/test/registered/unit/model_executor/test_num_token_non_padded_localization.py @@ -13,7 +13,7 @@ tensor and host-int twin agree, so a change to the sharding math fails loudly. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/model_executor/test_pool_configurator.py b/test/registered/unit/model_executor/test_pool_configurator.py index 26fc85b22a..2580f4465a 100644 --- a/test/registered/unit/model_executor/test_pool_configurator.py +++ b/test/registered/unit/model_executor/test_pool_configurator.py @@ -20,7 +20,7 @@ from sglang.srt.runtime_context import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") @contextlib.contextmanager diff --git a/test/registered/unit/model_executor/test_prefill_cuda_graph_runner.py b/test/registered/unit/model_executor/test_prefill_cuda_graph_runner.py index 29805fb3f2..28bc15f2ef 100644 --- a/test/registered/unit/model_executor/test_prefill_cuda_graph_runner.py +++ b/test/registered/unit/model_executor/test_prefill_cuda_graph_runner.py @@ -28,7 +28,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _FakeAttentionBackend: diff --git a/test/registered/unit/model_executor/test_prefill_cuda_graph_runner_helpers.py b/test/registered/unit/model_executor/test_prefill_cuda_graph_runner_helpers.py index 3d4ef6b89d..ac89658406 100644 --- a/test/registered/unit/model_executor/test_prefill_cuda_graph_runner_helpers.py +++ b/test/registered/unit/model_executor/test_prefill_cuda_graph_runner_helpers.py @@ -24,7 +24,7 @@ from sglang.srt.model_loader.utils import resolve_language_model from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _LayerModel: diff --git a/test/registered/unit/model_executor/test_unified_out_cache_loc_rebind.py b/test/registered/unit/model_executor/test_unified_out_cache_loc_rebind.py index 4cc9b006e5..a369950f3e 100644 --- a/test/registered/unit/model_executor/test_unified_out_cache_loc_rebind.py +++ b/test/registered/unit/model_executor/test_unified_out_cache_loc_rebind.py @@ -36,7 +36,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch, ForwardMo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _DEV = "cpu" diff --git a/test/registered/unit/model_loader/test_default_model_loader.py b/test/registered/unit/model_loader/test_default_model_loader.py index da5b3676ae..011a4648db 100644 --- a/test/registered/unit/model_loader/test_default_model_loader.py +++ b/test/registered/unit/model_loader/test_default_model_loader.py @@ -10,7 +10,7 @@ from sglang.srt.model_loader.loader import DefaultModelLoader from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestDefaultModelLoader(CustomTestCase): diff --git a/test/registered/unit/model_loader/test_modelopt_loader.py b/test/registered/unit/model_loader/test_modelopt_loader.py index 5bd16c7413..196f1cda3e 100644 --- a/test/registered/unit/model_loader/test_modelopt_loader.py +++ b/test/registered/unit/model_loader/test_modelopt_loader.py @@ -59,7 +59,7 @@ CALIBRATION_BATCH_SIZE = 36 CALIBRATION_NUM_SAMPLES = 512 DEFAULT_DEVICE = "cuda:0" -register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") class TestModelOptModelLoader(CustomTestCase): diff --git a/test/registered/unit/model_loader/test_prefetch_checkpoints.py b/test/registered/unit/model_loader/test_prefetch_checkpoints.py index c2f4439d23..f62158653e 100644 --- a/test/registered/unit/model_loader/test_prefetch_checkpoints.py +++ b/test/registered/unit/model_loader/test_prefetch_checkpoints.py @@ -29,7 +29,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _InlineThread: diff --git a/test/registered/unit/model_loader/test_presharded_loader.py b/test/registered/unit/model_loader/test_presharded_loader.py index f10a44d0d0..b9e5e82da5 100644 --- a/test/registered/unit/model_loader/test_presharded_loader.py +++ b/test/registered/unit/model_loader/test_presharded_loader.py @@ -18,7 +18,7 @@ from sglang.srt.model_loader.loader import PreshardedModelLoader from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestPreshardedHashTensor(unittest.TestCase): diff --git a/test/registered/unit/model_loader/test_remote_instance_loader.py b/test/registered/unit/model_loader/test_remote_instance_loader.py index 5962ccb994..df6cde5430 100644 --- a/test/registered/unit/model_loader/test_remote_instance_loader.py +++ b/test/registered/unit/model_loader/test_remote_instance_loader.py @@ -10,7 +10,7 @@ from sglang.srt.model_loader.remote_instance_weight_loader_utils import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") EXTRA_CONFIG = {"enable_multithread_load": True, "num_threads": 64} diff --git a/test/registered/unit/model_loader/test_runai_model_streamer_loader.py b/test/registered/unit/model_loader/test_runai_model_streamer_loader.py index 6554452174..6069773568 100644 --- a/test/registered/unit/model_loader/test_runai_model_streamer_loader.py +++ b/test/registered/unit/model_loader/test_runai_model_streamer_loader.py @@ -27,7 +27,7 @@ from sglang.srt.utils import runai_utils from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakeModel: diff --git a/test/registered/unit/model_loader/test_transformers_fallback.py b/test/registered/unit/model_loader/test_transformers_fallback.py index c18d20dd04..bcadb4d505 100644 --- a/test/registered/unit/model_loader/test_transformers_fallback.py +++ b/test/registered/unit/model_loader/test_transformers_fallback.py @@ -9,7 +9,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestTransformersFallbackWeightMapper(CustomTestCase): diff --git a/test/registered/unit/model_loader/test_weight_cache_protocol.py b/test/registered/unit/model_loader/test_weight_cache_protocol.py index f7e8d6b1c2..57bac7fa63 100644 --- a/test/registered/unit/model_loader/test_weight_cache_protocol.py +++ b/test/registered/unit/model_loader/test_weight_cache_protocol.py @@ -50,7 +50,7 @@ from sglang.srt.weight_cache.transport import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_cache_config(**overrides) -> CacheConfig: diff --git a/test/registered/unit/model_loader/test_weight_utils.py b/test/registered/unit/model_loader/test_weight_utils.py index 9ebaa72037..1721579ac7 100644 --- a/test/registered/unit/model_loader/test_weight_utils.py +++ b/test/registered/unit/model_loader/test_weight_utils.py @@ -15,7 +15,7 @@ from sglang.srt.utils import runai_utils from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") INDEX_NAME = "model.safetensors.index.json" diff --git a/test/registered/unit/models/test_bailing_vl_loader.py b/test/registered/unit/models/test_bailing_vl_loader.py index 8403c42738..dc1e1e30a9 100644 --- a/test/registered/unit/models/test_bailing_vl_loader.py +++ b/test/registered/unit/models/test_bailing_vl_loader.py @@ -16,7 +16,7 @@ from sglang.srt.models.bailing_mm_v3 import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _OneShotWeights: diff --git a/test/registered/unit/models/test_cosmos3.py b/test/registered/unit/models/test_cosmos3.py index de8b3b932c..ca5511f99d 100644 --- a/test/registered/unit/models/test_cosmos3.py +++ b/test/registered/unit/models/test_cosmos3.py @@ -16,7 +16,7 @@ All of this is pure CPU logic (no server / engine launch). from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") import os import tempfile diff --git a/test/registered/unit/models/test_cosmos3_edge.py b/test/registered/unit/models/test_cosmos3_edge.py index 539cf5a95c..835e1e3a97 100644 --- a/test/registered/unit/models/test_cosmos3_edge.py +++ b/test/registered/unit/models/test_cosmos3_edge.py @@ -15,7 +15,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import asyncio import unittest diff --git a/test/registered/unit/models/test_deepseek_mla_dispatch.py b/test/registered/unit/models/test_deepseek_mla_dispatch.py index dbd1448574..ef24631d7f 100644 --- a/test/registered/unit/models/test_deepseek_mla_dispatch.py +++ b/test/registered/unit/models/test_deepseek_mla_dispatch.py @@ -24,7 +24,7 @@ from sglang.srt.models.deepseek_common.attention_forward_methods.forward_methods from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def _fake_forward_batch(is_decode: bool): diff --git a/test/registered/unit/models/test_deepseek_nextn_mm_embed.py b/test/registered/unit/models/test_deepseek_nextn_mm_embed.py index dad4ab6dd6..162317069b 100644 --- a/test/registered/unit/models/test_deepseek_nextn_mm_embed.py +++ b/test/registered/unit/models/test_deepseek_nextn_mm_embed.py @@ -15,7 +15,7 @@ from sglang.srt.managers.schedule_batch import MM_PAD_SHIFT_VALUE from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=30, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") VOCAB_SIZE = 154880 HIDDEN_SIZE = 64 # tiny for CPU test diff --git a/test/registered/unit/models/test_deepseek_pp_mtp_embedding.py b/test/registered/unit/models/test_deepseek_pp_mtp_embedding.py index cfb5836701..fea10964b0 100644 --- a/test/registered/unit/models/test_deepseek_pp_mtp_embedding.py +++ b/test/registered/unit/models/test_deepseek_pp_mtp_embedding.py @@ -10,7 +10,7 @@ from sglang.srt.models.deepseek_v2 import pp_stage_needs_embedding from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _pp_group(*, rank: int, size: int) -> SimpleNamespace: diff --git a/test/registered/unit/models/test_deepseek_v4_amd_fused_mhc.py b/test/registered/unit/models/test_deepseek_v4_amd_fused_mhc.py index 52378f5b62..f4e785b322 100644 --- a/test/registered/unit/models/test_deepseek_v4_amd_fused_mhc.py +++ b/test/registered/unit/models/test_deepseek_v4_amd_fused_mhc.py @@ -6,7 +6,7 @@ from sglang.srt.models.deepseek_common.amd import deepseek_v4_fused_mhc from sglang.srt.runtime_context import override_platform from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestAmdFusedMhcCrossLayerGating(unittest.TestCase): diff --git a/test/registered/unit/models/test_deepseek_v4_mxfp4_shared_expert_requant.py b/test/registered/unit/models/test_deepseek_v4_mxfp4_shared_expert_requant.py index 88ecb9e9a6..a3ab25de1c 100644 --- a/test/registered/unit/models/test_deepseek_v4_mxfp4_shared_expert_requant.py +++ b/test/registered/unit/models/test_deepseek_v4_mxfp4_shared_expert_requant.py @@ -6,7 +6,7 @@ from sglang.srt.layers.quantization.fp8_utils import quantize_block_fp8_weight_t from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _E2M1_LUT = torch.tensor([0.0, 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, 6.0]) diff --git a/test/registered/unit/models/test_deepseek_v4_rope_policy.py b/test/registered/unit/models/test_deepseek_v4_rope_policy.py index bae5e90ec2..bd12b3f967 100644 --- a/test/registered/unit/models/test_deepseek_v4_rope_policy.py +++ b/test/registered/unit/models/test_deepseek_v4_rope_policy.py @@ -13,7 +13,7 @@ from sglang.srt.environ import envs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _ModuleStub(nn.Module): diff --git a/test/registered/unit/models/test_deepseek_v4_shared_expert_fusion.py b/test/registered/unit/models/test_deepseek_v4_shared_expert_fusion.py index fd47b414d9..18ac0fd522 100644 --- a/test/registered/unit/models/test_deepseek_v4_shared_expert_fusion.py +++ b/test/registered/unit/models/test_deepseek_v4_shared_expert_fusion.py @@ -16,7 +16,7 @@ from sglang.srt.runtime_context import get_context, get_exec, get_flags, get_par from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestDeepseekV4SharedExpertFusionPolicy(CustomTestCase): diff --git a/test/registered/unit/models/test_draft_entry_hook_parity.py b/test/registered/unit/models/test_draft_entry_hook_parity.py index d1ebb4ce2e..7dafab1d74 100644 --- a/test/registered/unit/models/test_draft_entry_hook_parity.py +++ b/test/registered/unit/models/test_draft_entry_hook_parity.py @@ -22,7 +22,7 @@ from sglang.srt.models.registry import ModelRegistry from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") # The hooks the loader resolves on the entry class where a draft/target # disagreement is a defect rather than a difference. Weight-name maps diff --git a/test/registered/unit/models/test_glm5_next_bfg_fusion.py b/test/registered/unit/models/test_glm5_next_bfg_fusion.py index d3b3411a55..78a2ce9f5d 100644 --- a/test/registered/unit/models/test_glm5_next_bfg_fusion.py +++ b/test/registered/unit/models/test_glm5_next_bfg_fusion.py @@ -10,7 +10,7 @@ from sglang.srt.models import glm5_next from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") PREFIX = "model.layers.0.self_attn" QKV = ("q_proj", "k_proj", "v_proj") diff --git a/test/registered/unit/models/test_glm5_next_modelopt.py b/test/registered/unit/models/test_glm5_next_modelopt.py index 459805680d..663593d3f6 100644 --- a/test/registered/unit/models/test_glm5_next_modelopt.py +++ b/test/registered/unit/models/test_glm5_next_modelopt.py @@ -9,7 +9,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") NVIDIA_EXCLUDE_MODULES = [ "model.language_model.embed_tokens", diff --git a/test/registered/unit/models/test_glm_moe_gate_fp32.py b/test/registered/unit/models/test_glm_moe_gate_fp32.py index f80c0a1045..8e03ebbcca 100644 --- a/test/registered/unit/models/test_glm_moe_gate_fp32.py +++ b/test/registered/unit/models/test_glm_moe_gate_fp32.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/models/test_glm_nextn_moe_ptpc.py b/test/registered/unit/models/test_glm_nextn_moe_ptpc.py index e047cbf502..97148edb4d 100644 --- a/test/registered/unit/models/test_glm_nextn_moe_ptpc.py +++ b/test/registered/unit/models/test_glm_nextn_moe_ptpc.py @@ -24,7 +24,7 @@ from sglang.srt.models.glm4_moe import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") LAYER = 78 PREFIX = f"model.layers.{LAYER}" diff --git a/test/registered/unit/models/test_glmmoedsa_correction_bias_fp32.py b/test/registered/unit/models/test_glmmoedsa_correction_bias_fp32.py index bc958e48b9..42c1ffd6c3 100644 --- a/test/registered/unit/models/test_glmmoedsa_correction_bias_fp32.py +++ b/test/registered/unit/models/test_glmmoedsa_correction_bias_fp32.py @@ -20,7 +20,7 @@ from sglang.srt.models.deepseek_v2 import MoEGate from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") GLM_MAIN_ARCH = "GlmMoeDsaForCausalLM" GLM_NEXTN_ARCH = "GlmMoeDsaForCausalLMNextN" # draft head, rewritten in model_config.py diff --git a/test/registered/unit/models/test_gpt_oss_runai_ownership.py b/test/registered/unit/models/test_gpt_oss_runai_ownership.py index add513377f..54dd81d9e4 100644 --- a/test/registered/unit/models/test_gpt_oss_runai_ownership.py +++ b/test/registered/unit/models/test_gpt_oss_runai_ownership.py @@ -18,7 +18,7 @@ from sglang.srt.models.gpt_oss import GptOssForCausalLM from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _streamed(value: float) -> torch.Tensor: diff --git a/test/registered/unit/models/test_granitemoe_split_expert_loading.py b/test/registered/unit/models/test_granitemoe_split_expert_loading.py index 730b40d4b5..e7bacc6675 100644 --- a/test/registered/unit/models/test_granitemoe_split_expert_loading.py +++ b/test/registered/unit/models/test_granitemoe_split_expert_loading.py @@ -11,7 +11,7 @@ silently dropped: the server started normally and then emitted garbage from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/models/test_hunyuan_v3_nextn_weight_loading.py b/test/registered/unit/models/test_hunyuan_v3_nextn_weight_loading.py index 8a5eb099c8..bd8b68e8c2 100644 --- a/test/registered/unit/models/test_hunyuan_v3_nextn_weight_loading.py +++ b/test/registered/unit/models/test_hunyuan_v3_nextn_weight_loading.py @@ -9,7 +9,7 @@ being remapped to a nonexistent decoder parameter and silently dropped. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/models/test_hunyuan_v4.py b/test/registered/unit/models/test_hunyuan_v4.py index a9069ee81e..ddc8c8a92c 100644 --- a/test/registered/unit/models/test_hunyuan_v4.py +++ b/test/registered/unit/models/test_hunyuan_v4.py @@ -9,7 +9,7 @@ from sglang.srt.models import hunyuan_v4 from sglang.srt.models.deepseek_common.attention_forward_methods import forward_mla from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def test_attention_gate_uses_attention_tp(monkeypatch): diff --git a/test/registered/unit/models/test_hunyuan_v4_nextn_weight_loading.py b/test/registered/unit/models/test_hunyuan_v4_nextn_weight_loading.py index da3a16c45b..c5df9ef120 100644 --- a/test/registered/unit/models/test_hunyuan_v4_nextn_weight_loading.py +++ b/test/registered/unit/models/test_hunyuan_v4_nextn_weight_loading.py @@ -6,7 +6,7 @@ import torch from sglang.srt.models.hunyuan_v4_nextn import HYV4ForCausalLMNextN from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHunyuanV4NextNWeightLoading(unittest.TestCase): diff --git a/test/registered/unit/models/test_interns1pro_processor.py b/test/registered/unit/models/test_interns1pro_processor.py index 36009bc9ea..d462bda2a2 100644 --- a/test/registered/unit/models/test_interns1pro_processor.py +++ b/test/registered/unit/models/test_interns1pro_processor.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") from types import SimpleNamespace from unittest.mock import Mock diff --git a/test/registered/unit/models/test_kimi_k25.py b/test/registered/unit/models/test_kimi_k25.py index 4d79380bf5..797522f7b5 100644 --- a/test/registered/unit/models/test_kimi_k25.py +++ b/test/registered/unit/models/test_kimi_k25.py @@ -76,7 +76,7 @@ from sglang.srt.server_args import ServerArgs from sglang.srt.utils import ImageData from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _MoonViT3dTower: diff --git a/test/registered/unit/models/test_kimi_k3_vision.py b/test/registered/unit/models/test_kimi_k3_vision.py index bf9e6ad08f..d8f4776345 100644 --- a/test/registered/unit/models/test_kimi_k3_vision.py +++ b/test/registered/unit/models/test_kimi_k3_vision.py @@ -24,7 +24,7 @@ from sglang.srt.multimodal.mm_utils import run_dp_sharded_mrope_vision_model from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") @pytest.mark.parametrize( diff --git a/test/registered/unit/models/test_kimi_vl.py b/test/registered/unit/models/test_kimi_vl.py index 1cd09aa7d5..8a5d1afbf0 100644 --- a/test/registered/unit/models/test_kimi_vl.py +++ b/test/registered/unit/models/test_kimi_vl.py @@ -20,7 +20,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class _VisionTower: diff --git a/test/registered/unit/models/test_kimi_vl_moonvit.py b/test/registered/unit/models/test_kimi_vl_moonvit.py index dffe24e4d3..af2f34c177 100644 --- a/test/registered/unit/models/test_kimi_vl_moonvit.py +++ b/test/registered/unit/models/test_kimi_vl_moonvit.py @@ -3,7 +3,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") from sglang.srt.models import kimi_vl_moonvit from sglang.srt.models.kimi_vl_moonvit import Learnable2DInterpPosEmb diff --git a/test/registered/unit/models/test_llava.py b/test/registered/unit/models/test_llava.py index 6f981a3c2b..011b267738 100644 --- a/test/registered/unit/models/test_llava.py +++ b/test/registered/unit/models/test_llava.py @@ -8,7 +8,7 @@ from sglang.srt.multimodal.processors.llava import LlavaImageProcessor from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class PixtralVisionConfig: diff --git a/test/registered/unit/models/test_llava_processor_pool.py b/test/registered/unit/models/test_llava_processor_pool.py index 32623c4d1a..56e0932dbc 100644 --- a/test/registered/unit/models/test_llava_processor_pool.py +++ b/test/registered/unit/models/test_llava_processor_pool.py @@ -10,7 +10,7 @@ import pytest from sglang.srt.multimodal.processors.llava import LlavaImageProcessor from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class _BrokenExecutor(concurrent.futures.Executor): diff --git a/test/registered/unit/models/test_locate_anything.py b/test/registered/unit/models/test_locate_anything.py index c870fd2184..785a378f07 100644 --- a/test/registered/unit/models/test_locate_anything.py +++ b/test/registered/unit/models/test_locate_anything.py @@ -19,7 +19,7 @@ from sglang.srt.models.locate_anything import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _small_config(): diff --git a/test/registered/unit/models/test_longcat_flash_router_hpc_gemm.py b/test/registered/unit/models/test_longcat_flash_router_hpc_gemm.py index 754263be1e..d871730ca0 100644 --- a/test/registered/unit/models/test_longcat_flash_router_hpc_gemm.py +++ b/test/registered/unit/models/test_longcat_flash_router_hpc_gemm.py @@ -13,7 +13,7 @@ maybe_stub_sgl_kernel() from sglang.srt.models.longcat_flash import LongcatFlashRouter # noqa: E402 -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _longcat_config(hidden_size, n_routed_experts, *, router_bias=False): diff --git a/test/registered/unit/models/test_mellum.py b/test/registered/unit/models/test_mellum.py index 112c43f758..208e53c12f 100644 --- a/test/registered/unit/models/test_mellum.py +++ b/test/registered/unit/models/test_mellum.py @@ -8,7 +8,7 @@ from sglang.srt.models.mellum import MellumForCausalLM from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestMellumForCausalLM(CustomTestCase): diff --git a/test/registered/unit/models/test_mistral3_vision_feature.py b/test/registered/unit/models/test_mistral3_vision_feature.py index c7df5d06b0..b010e7d427 100644 --- a/test/registered/unit/models/test_mistral3_vision_feature.py +++ b/test/registered/unit/models/test_mistral3_vision_feature.py @@ -15,7 +15,7 @@ from sglang.srt.models.llava import LlavaForConditionalGeneration from sglang.srt.models.mistral import Mistral3ForConditionalGeneration from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=29, suite="base-a-test-cpu") NUM_LAYERS = 4 TOKENS = 3 diff --git a/test/registered/unit/models/test_moss_vl_processor.py b/test/registered/unit/models/test_moss_vl_processor.py index 04c14ff8d7..368b4e9edd 100644 --- a/test/registered/unit/models/test_moss_vl_processor.py +++ b/test/registered/unit/models/test_moss_vl_processor.py @@ -10,7 +10,7 @@ import torch from sglang.srt.multimodal.processors.moss_vl import MossVLImageProcessor from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _vision_info(frame_count: int): diff --git a/test/registered/unit/models/test_nano_nemotron_vl.py b/test/registered/unit/models/test_nano_nemotron_vl.py index 5a31c35360..63309f9dca 100644 --- a/test/registered/unit/models/test_nano_nemotron_vl.py +++ b/test/registered/unit/models/test_nano_nemotron_vl.py @@ -13,7 +13,7 @@ from sglang.srt.models.nano_nemotron_vl import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestNemotronHOmniModel(CustomTestCase): diff --git a/test/registered/unit/models/test_nemotron_h_mtp.py b/test/registered/unit/models/test_nemotron_h_mtp.py index d6de48d818..35f8e62cf8 100644 --- a/test/registered/unit/models/test_nemotron_h_mtp.py +++ b/test/registered/unit/models/test_nemotron_h_mtp.py @@ -18,7 +18,7 @@ from sglang.srt.models.nemotron_h_mtp import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class _RecordingLayer(nn.Module): diff --git a/test/registered/unit/models/test_nemotron_h_shared_add.py b/test/registered/unit/models/test_nemotron_h_shared_add.py index 08d747bf45..5adb576f91 100644 --- a/test/registered/unit/models/test_nemotron_h_shared_add.py +++ b/test/registered/unit/models/test_nemotron_h_shared_add.py @@ -8,7 +8,7 @@ These cases pin the gate that decides when the substitution is safe. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=20, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/models/test_nemotron_h_weight_loading.py b/test/registered/unit/models/test_nemotron_h_weight_loading.py index 1943cd1fcd..19560a4d5d 100644 --- a/test/registered/unit/models/test_nemotron_h_weight_loading.py +++ b/test/registered/unit/models/test_nemotron_h_weight_loading.py @@ -10,7 +10,7 @@ from sglang.srt.models.nemotron_h_mtp import NemotronHForCausalLMMTP from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _FakePPGroup: diff --git a/test/registered/unit/models/test_paddleocr_vl_vision.py b/test/registered/unit/models/test_paddleocr_vl_vision.py index 91189293c2..5e57433e72 100644 --- a/test/registered/unit/models/test_paddleocr_vl_vision.py +++ b/test/registered/unit/models/test_paddleocr_vl_vision.py @@ -18,7 +18,7 @@ from sglang.srt.models.paddleocr_vl import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") # Mixes repeated grids (LFU cache hits), an odd aspect ratio, and t > 1. GRIDS = [(1, 4, 6), (1, 8, 10), (2, 2, 4), (1, 8, 10)] diff --git a/test/registered/unit/models/test_qwen3_5_modelopt_fp4.py b/test/registered/unit/models/test_qwen3_5_modelopt_fp4.py index d0e3f60ad9..8132152acc 100644 --- a/test/registered/unit/models/test_qwen3_5_modelopt_fp4.py +++ b/test/registered/unit/models/test_qwen3_5_modelopt_fp4.py @@ -21,7 +21,7 @@ from sglang.srt.models.qwen3_5 import QWEN3_5_KV_SCALE_MAPPER from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestModelOptFp4AttentionExclusion(CustomTestCase): diff --git a/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py b/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py index 57aa6457fa..a748fbf497 100644 --- a/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py +++ b/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py @@ -10,7 +10,7 @@ Regression test for https://github.com/sgl-project/sglang/issues/23051 from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py b/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py index 402df38e78..9a9116f086 100644 --- a/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py +++ b/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py @@ -9,7 +9,7 @@ from sglang.srt.models.qwen3_5_mtp import Qwen3_5ForCausalLMMTP from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestQwen3_5PipelineParallel(CustomTestCase): diff --git a/test/registered/unit/models/test_qwen3_embedding_registration.py b/test/registered/unit/models/test_qwen3_embedding_registration.py index 3f11f3c81c..f79d8a584a 100644 --- a/test/registered/unit/models/test_qwen3_embedding_registration.py +++ b/test/registered/unit/models/test_qwen3_embedding_registration.py @@ -8,7 +8,7 @@ and be served as an embedding model, NOT fall back to the Transformers backend. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=17, suite="base-a-test-cpu") +register_cpu_ci(est_time=11, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/models/test_qwen3_vl_feature_materialization.py b/test/registered/unit/models/test_qwen3_vl_feature_materialization.py index 34262c1adb..4d0191d6dd 100644 --- a/test/registered/unit/models/test_qwen3_vl_feature_materialization.py +++ b/test/registered/unit/models/test_qwen3_vl_feature_materialization.py @@ -24,7 +24,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _RecordingVisual: diff --git a/test/registered/unit/models/test_qwen4_exp_ple_table.py b/test/registered/unit/models/test_qwen4_exp_ple_table.py index 81063c33fc..6bed0d6324 100644 --- a/test/registered/unit/models/test_qwen4_exp_ple_table.py +++ b/test/registered/unit/models/test_qwen4_exp_ple_table.py @@ -34,7 +34,7 @@ from sglang.srt.models.qwen4_exp_ple_table import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class TestPleFileTableAllocator(CustomTestCase): diff --git a/test/registered/unit/models/test_radio.py b/test/registered/unit/models/test_radio.py index 3ef21eccba..08bcd28b71 100644 --- a/test/registered/unit/models/test_radio.py +++ b/test/registered/unit/models/test_radio.py @@ -10,7 +10,7 @@ from sglang.srt.models.radio import RadioModel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=3, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _RecordingWeight: diff --git a/test/registered/unit/models/test_rocm_hisparse_fused_kv.py b/test/registered/unit/models/test_rocm_hisparse_fused_kv.py index e5c49e6529..10cb95655b 100644 --- a/test/registered/unit/models/test_rocm_hisparse_fused_kv.py +++ b/test/registered/unit/models/test_rocm_hisparse_fused_kv.py @@ -12,7 +12,7 @@ from sglang.srt.models.deepseek_common.attention_forward_methods import forward_ from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestRocmHiSparseFusedKV(CustomTestCase): diff --git a/test/registered/unit/models/test_shared_experts_fusion_gates.py b/test/registered/unit/models/test_shared_experts_fusion_gates.py index 96f6dbe4bb..215e2209e3 100644 --- a/test/registered/unit/models/test_shared_experts_fusion_gates.py +++ b/test/registered/unit/models/test_shared_experts_fusion_gates.py @@ -26,7 +26,7 @@ from sglang.srt.runtime_context import get_context, get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") def _quant(name: str): diff --git a/test/registered/unit/models/test_xllm.py b/test/registered/unit/models/test_xllm.py index 90e9e563ab..a516342e6c 100644 --- a/test/registered/unit/models/test_xllm.py +++ b/test/registered/unit/models/test_xllm.py @@ -35,7 +35,7 @@ from sglang.srt.runtime_context import get_context from sglang.srt.utils.hf_transformers.common import _CONFIG_REGISTRY from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class _IdentityRotary: diff --git a/test/registered/unit/models/test_zaya_cca.py b/test/registered/unit/models/test_zaya_cca.py index c4d2c74983..21c0c1dac2 100644 --- a/test/registered/unit/models/test_zaya_cca.py +++ b/test/registered/unit/models/test_zaya_cca.py @@ -34,7 +34,7 @@ from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.layer_ut_utils import init_single_process_dist from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _ensure_dist_initialized(cls) -> None: diff --git a/test/registered/unit/models/test_zaya_mod_tp.py b/test/registered/unit/models/test_zaya_mod_tp.py index d5d45d6060..91e6dab2b6 100644 --- a/test/registered/unit/models/test_zaya_mod_tp.py +++ b/test/registered/unit/models/test_zaya_mod_tp.py @@ -25,7 +25,7 @@ from sglang.srt.models.zaya import mod_blend, mod_premask_experts from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _reference_blend( diff --git a/test/registered/unit/multimodal/rust/inkling/test_bindings.py b/test/registered/unit/multimodal/rust/inkling/test_bindings.py index 06893e5a20..0a4fb17b24 100644 --- a/test/registered/unit/multimodal/rust/inkling/test_bindings.py +++ b/test/registered/unit/multimodal/rust/inkling/test_bindings.py @@ -22,7 +22,7 @@ from sglang.test.test_utils import CustomTestCase sys.path.insert(0, str(Path(__file__).resolve().parents[1])) from _mm_rust_utils import image_bytes, load_core # noqa: E402 -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") CORE = load_core() PATCH_SIZE = 16 diff --git a/test/registered/unit/multimodal/rust/qwen/test_driver.py b/test/registered/unit/multimodal/rust/qwen/test_driver.py index efd03b2ed7..c86c6583d3 100644 --- a/test/registered/unit/multimodal/rust/qwen/test_driver.py +++ b/test/registered/unit/multimodal/rust/qwen/test_driver.py @@ -33,7 +33,7 @@ from _mm_rust_utils import ( # noqa: E402 spec_json, ) -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") QWEN_CORE = getattr(load_core(), "qwen_vl", None) SPEC = spec_json(PROCESSOR_CONFIGS["qwen2_5_vl"]) diff --git a/test/registered/unit/multimodal/rust/qwen/test_e2e_parity.py b/test/registered/unit/multimodal/rust/qwen/test_e2e_parity.py index c9531036a3..21b4a82911 100644 --- a/test/registered/unit/multimodal/rust/qwen/test_e2e_parity.py +++ b/test/registered/unit/multimodal/rust/qwen/test_e2e_parity.py @@ -30,7 +30,7 @@ sys.path.insert(0, str(Path(__file__).resolve().parents[1])) from _fixtures import make_processor, snapshot # noqa: E402 from _mm_rust_utils import PROCESSOR_CONFIGS, image_bytes, load_core # noqa: E402 -register_cpu_ci(est_time=16, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") CORE = load_core() DRIVER = getattr(getattr(CORE, "qwen_vl", None), "process_mm", None) diff --git a/test/registered/unit/multimodal/rust/qwen/test_preprocess.py b/test/registered/unit/multimodal/rust/qwen/test_preprocess.py index f93222aff3..bb1e0f0885 100644 --- a/test/registered/unit/multimodal/rust/qwen/test_preprocess.py +++ b/test/registered/unit/multimodal/rust/qwen/test_preprocess.py @@ -32,7 +32,7 @@ from _mm_rust_utils import ( # noqa: E402 spec_json, ) -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") QWEN_CORE = getattr(load_core(), "qwen_vl", None) diff --git a/test/registered/unit/multimodal/rust/qwen/test_rust_mm_processor.py b/test/registered/unit/multimodal/rust/qwen/test_rust_mm_processor.py index b0e5afc716..fd5b40d243 100644 --- a/test/registered/unit/multimodal/rust/qwen/test_rust_mm_processor.py +++ b/test/registered/unit/multimodal/rust/qwen/test_rust_mm_processor.py @@ -29,7 +29,7 @@ sys.path.insert(0, str(Path(__file__).resolve().parents[1])) from _fixtures import make_processor # noqa: E402 from _mm_rust_utils import PROCESSOR_CONFIGS, image_bytes, load_core # noqa: E402 -register_cpu_ci(est_time=16, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") CORE = load_core() DRIVER = getattr(getattr(CORE, "qwen_vl", None), "process_mm", None) diff --git a/test/registered/unit/multimodal/rust/qwen/test_token_layout_mrope.py b/test/registered/unit/multimodal/rust/qwen/test_token_layout_mrope.py index c3e0796e11..c8bf0396ff 100644 --- a/test/registered/unit/multimodal/rust/qwen/test_token_layout_mrope.py +++ b/test/registered/unit/multimodal/rust/qwen/test_token_layout_mrope.py @@ -30,7 +30,7 @@ from _mm_rust_utils import ( # noqa: E402 spec_json, ) -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") QWEN_CORE = getattr(load_core(), "qwen_vl", None) diff --git a/test/registered/unit/multimodal/rust/shared/test_fetch.py b/test/registered/unit/multimodal/rust/shared/test_fetch.py index 996ff53041..6c3b6e2072 100644 --- a/test/registered/unit/multimodal/rust/shared/test_fetch.py +++ b/test/registered/unit/multimodal/rust/shared/test_fetch.py @@ -20,7 +20,7 @@ from sglang.test.test_utils import CustomTestCase sys.path.insert(0, str(Path(__file__).resolve().parents[1])) from _mm_rust_utils import load_core # noqa: E402 -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") CORE = load_core() FETCH = CORE and CORE.common.fetch_bytes diff --git a/test/registered/unit/multimodal/rust/shared/test_image_decode.py b/test/registered/unit/multimodal/rust/shared/test_image_decode.py index fd174f8604..bd87503f52 100644 --- a/test/registered/unit/multimodal/rust/shared/test_image_decode.py +++ b/test/registered/unit/multimodal/rust/shared/test_image_decode.py @@ -19,7 +19,7 @@ from sglang.test.test_utils import CustomTestCase sys.path.insert(0, str(Path(__file__).resolve().parents[1])) from _mm_rust_utils import load_core # noqa: E402 -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") CORE = load_core() DECODE = CORE and CORE.common.image_decode_rgb diff --git a/test/registered/unit/multimodal/rust/shared/test_partition_cores.py b/test/registered/unit/multimodal/rust/shared/test_partition_cores.py index eb33f10266..4a97e625c7 100644 --- a/test/registered/unit/multimodal/rust/shared/test_partition_cores.py +++ b/test/registered/unit/multimodal/rust/shared/test_partition_cores.py @@ -14,7 +14,7 @@ maybe_stub_sgl_kernel() from sglang.srt.rust_server.config import _partition_cores # noqa: E402 -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def partition(node_cores, **kwargs): diff --git a/test/registered/unit/multimodal/rust/shared/test_rust_mm_gate.py b/test/registered/unit/multimodal/rust/shared/test_rust_mm_gate.py index 517a09b1c1..4db5255cd0 100644 --- a/test/registered/unit/multimodal/rust/shared/test_rust_mm_gate.py +++ b/test/registered/unit/multimodal/rust/shared/test_rust_mm_gate.py @@ -24,7 +24,7 @@ from sglang.srt.runtime_context import publish from sglang.srt.rust_server.multimodal import rust_mm_family_for # noqa: E402 from sglang.srt.server_args import ServerArgs -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") def processor_cls_for(architecture, model_type): diff --git a/test/registered/unit/multimodal/rust/shared/test_rust_server_extension.py b/test/registered/unit/multimodal/rust/shared/test_rust_server_extension.py index 42bb14029b..8589501b75 100644 --- a/test/registered/unit/multimodal/rust/shared/test_rust_server_extension.py +++ b/test/registered/unit/multimodal/rust/shared/test_rust_server_extension.py @@ -12,7 +12,7 @@ maybe_stub_sgl_kernel() from sglang.srt.rust_server import server as server_module # noqa: E402 from sglang.srt.rust_server.server import RustServer # noqa: E402 -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestRustServerExtension(CustomTestCase): diff --git a/test/registered/unit/multimodal/rust/shared/test_wrap_encoded.py b/test/registered/unit/multimodal/rust/shared/test_wrap_encoded.py index a11821de4d..00e72ddfaf 100644 --- a/test/registered/unit/multimodal/rust/shared/test_wrap_encoded.py +++ b/test/registered/unit/multimodal/rust/shared/test_wrap_encoded.py @@ -20,7 +20,7 @@ from sglang.srt.rust_server.multimodal import ( # noqa: E402 RustMmSpec, ) -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class TestWrapEncoded(CustomTestCase): diff --git a/test/registered/unit/multimodal/test_audio_container_decode.py b/test/registered/unit/multimodal/test_audio_container_decode.py index 2c65e329eb..2ec8b7a61c 100644 --- a/test/registered/unit/multimodal/test_audio_container_decode.py +++ b/test/registered/unit/multimodal/test_audio_container_decode.py @@ -24,7 +24,7 @@ from sglang.srt.utils.common import load_audio from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") def _tone(*, sample_rate: int, channels: int = 1) -> np.ndarray: diff --git a/test/registered/unit/multimodal/test_bailing_mm_processor.py b/test/registered/unit/multimodal/test_bailing_mm_processor.py index 803d9dda6e..dcd28f76af 100644 --- a/test/registered/unit/multimodal/test_bailing_mm_processor.py +++ b/test/registered/unit/multimodal/test_bailing_mm_processor.py @@ -10,7 +10,7 @@ from sglang.srt.multimodal.processors.bailing_mm import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestBailingMMProcessor(CustomTestCase): diff --git a/test/registered/unit/multimodal/test_base_processor_bad_input.py b/test/registered/unit/multimodal/test_base_processor_bad_input.py index c66a3f9b51..d13be95381 100644 --- a/test/registered/unit/multimodal/test_base_processor_bad_input.py +++ b/test/registered/unit/multimodal/test_base_processor_bad_input.py @@ -6,7 +6,7 @@ Media the client supplied but that cannot be fetched or decoded must raise from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import base64 import binascii diff --git a/test/registered/unit/multimodal/test_base_processor_image_decode.py b/test/registered/unit/multimodal/test_base_processor_image_decode.py index f6ca7a2235..4fb52f6079 100644 --- a/test/registered/unit/multimodal/test_base_processor_image_decode.py +++ b/test/registered/unit/multimodal/test_base_processor_image_decode.py @@ -11,7 +11,7 @@ No server, no model loading — pure CPU. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import asyncio import concurrent.futures diff --git a/test/registered/unit/multimodal/test_cuda_ipc_pool_budget.py b/test/registered/unit/multimodal/test_cuda_ipc_pool_budget.py index ad98e9f93a..b42253d9c4 100644 --- a/test/registered/unit/multimodal/test_cuda_ipc_pool_budget.py +++ b/test/registered/unit/multimodal/test_cuda_ipc_pool_budget.py @@ -7,7 +7,7 @@ from sglang.srt.multimodal.transport.cuda_ipc import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestCudaIpcPoolBudget(unittest.TestCase): diff --git a/test/registered/unit/multimodal/test_cuda_vmm_transport.py b/test/registered/unit/multimodal/test_cuda_vmm_transport.py index 5923d48362..134fe015fb 100644 --- a/test/registered/unit/multimodal/test_cuda_vmm_transport.py +++ b/test/registered/unit/multimodal/test_cuda_vmm_transport.py @@ -23,7 +23,7 @@ from sglang.srt.utils.cuda_vmm_transport_utils import ( from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=31, stage="base-c", runner_config="4-gpu-gb300") +register_cuda_ci(est_time=28, stage="base-c", runner_config="4-gpu-gb300") class _FabricUnavailableCudaVmmMemoryPool(CudaVmmMemoryPool): diff --git a/test/registered/unit/multimodal/test_deepseek_ocr_geometry.py b/test/registered/unit/multimodal/test_deepseek_ocr_geometry.py index 6dba0e043f..10bfe27ae6 100644 --- a/test/registered/unit/multimodal/test_deepseek_ocr_geometry.py +++ b/test/registered/unit/multimodal/test_deepseek_ocr_geometry.py @@ -25,7 +25,7 @@ from sglang.srt.multimodal.processors.deepseek_ocr import apply_ocr_geometry from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=60, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _hf_config(vision_model_name: str, projector_input_dim) -> SimpleNamespace: diff --git a/test/registered/unit/multimodal/test_dots_note_omni.py b/test/registered/unit/multimodal/test_dots_note_omni.py index 1fb58fb9f3..bf88dfbabc 100644 --- a/test/registered/unit/multimodal/test_dots_note_omni.py +++ b/test/registered/unit/multimodal/test_dots_note_omni.py @@ -16,7 +16,7 @@ from sglang.srt.multimodal.processors.dots_note_omni import DotsNoteOmniProcesso from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") _IM_TOKEN_ID = 100 _AUDIO_TOKEN_ID = 200 diff --git a/test/registered/unit/multimodal/test_evs.py b/test/registered/unit/multimodal/test_evs.py index 047901ff91..a777c0b3a7 100644 --- a/test/registered/unit/multimodal/test_evs.py +++ b/test/registered/unit/multimodal/test_evs.py @@ -7,7 +7,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import run_doctests -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def test_resolve_evs_config(): diff --git a/test/registered/unit/multimodal/test_feature_materialization.py b/test/registered/unit/multimodal/test_feature_materialization.py index 6fcc280125..02d460ede6 100644 --- a/test/registered/unit/multimodal/test_feature_materialization.py +++ b/test/registered/unit/multimodal/test_feature_materialization.py @@ -8,7 +8,7 @@ from sglang.srt.multimodal.mm_utils import materialize_multimodal_features from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestFeatureMaterialization(CustomTestCase): diff --git a/test/registered/unit/multimodal/test_glm4v_mixed_offsets.py b/test/registered/unit/multimodal/test_glm4v_mixed_offsets.py index 1384c12c16..9072bc3d87 100644 --- a/test/registered/unit/multimodal/test_glm4v_mixed_offsets.py +++ b/test/registered/unit/multimodal/test_glm4v_mixed_offsets.py @@ -6,7 +6,7 @@ from sglang.srt.multimodal.processors.base_processor import MultimodalSpecialTok from sglang.srt.multimodal.processors.glm4v import Glm4vImageProcessor from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=9, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _processor(): diff --git a/test/registered/unit/multimodal/test_gpu_feature_transport.py b/test/registered/unit/multimodal/test_gpu_feature_transport.py index 670e68a04e..6d6801c100 100644 --- a/test/registered/unit/multimodal/test_gpu_feature_transport.py +++ b/test/registered/unit/multimodal/test_gpu_feature_transport.py @@ -11,7 +11,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestCudaVmmFeatureTransport(unittest.TestCase): diff --git a/test/registered/unit/multimodal/test_kimi_k3_gpu_preprocess.py b/test/registered/unit/multimodal/test_kimi_k3_gpu_preprocess.py index f3988bd260..bfeb5fc7a1 100644 --- a/test/registered/unit/multimodal/test_kimi_k3_gpu_preprocess.py +++ b/test/registered/unit/multimodal/test_kimi_k3_gpu_preprocess.py @@ -11,7 +11,7 @@ from sglang.srt.multimodal.processors.kimi_k3 import _fill_transparent_bg from sglang.srt.multimodal.processors.kimi_k25 import _resize_bicubic_if_needed from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") def _natural_image(height: int, width: int) -> np.ndarray: diff --git a/test/registered/unit/multimodal/test_media_artifact_processor.py b/test/registered/unit/multimodal/test_media_artifact_processor.py index be4d8da6d8..02d77de429 100644 --- a/test/registered/unit/multimodal/test_media_artifact_processor.py +++ b/test/registered/unit/multimodal/test_media_artifact_processor.py @@ -17,7 +17,7 @@ from sglang.srt.multimodal.media_artifacts import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") @dataclass(frozen=True) diff --git a/test/registered/unit/multimodal/test_media_url_security.py b/test/registered/unit/multimodal/test_media_url_security.py index 1959bf6b8c..9e06e6be8c 100644 --- a/test/registered/unit/multimodal/test_media_url_security.py +++ b/test/registered/unit/multimodal/test_media_url_security.py @@ -16,7 +16,7 @@ from sglang.srt.utils.common import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") class _MediaHandler(http.server.BaseHTTPRequestHandler): diff --git a/test/registered/unit/multimodal/test_mrope_encoder_utils.py b/test/registered/unit/multimodal/test_mrope_encoder_utils.py index a493461f02..4ab29ab19d 100644 --- a/test/registered/unit/multimodal/test_mrope_encoder_utils.py +++ b/test/registered/unit/multimodal/test_mrope_encoder_utils.py @@ -13,7 +13,7 @@ from sglang.srt.runtime_context import get_parallel from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _RecordingGather: diff --git a/test/registered/unit/multimodal/test_nano_nemotron_vl_processor.py b/test/registered/unit/multimodal/test_nano_nemotron_vl_processor.py index cee8947432..3bab964c3f 100644 --- a/test/registered/unit/multimodal/test_nano_nemotron_vl_processor.py +++ b/test/registered/unit/multimodal/test_nano_nemotron_vl_processor.py @@ -9,7 +9,7 @@ from sglang.srt.multimodal.processors.nano_nemotron_vl import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestNanoNemotronVLProcessor(CustomTestCase): diff --git a/test/registered/unit/multimodal/test_pixtral_processor.py b/test/registered/unit/multimodal/test_pixtral_processor.py index 22b4d637e3..b45607a08b 100644 --- a/test/registered/unit/multimodal/test_pixtral_processor.py +++ b/test/registered/unit/multimodal/test_pixtral_processor.py @@ -16,7 +16,7 @@ from sglang.srt.multimodal.processors.pixtral import PixtralProcessor from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestPixtralProcessor(CustomTestCase): diff --git a/test/registered/unit/multimodal/test_precomputed_embedding_validation.py b/test/registered/unit/multimodal/test_precomputed_embedding_validation.py index a10c7147ea..0f85ce53a5 100644 --- a/test/registered/unit/multimodal/test_precomputed_embedding_validation.py +++ b/test/registered/unit/multimodal/test_precomputed_embedding_validation.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") import unittest diff --git a/test/registered/unit/multimodal/test_preprocess_cache.py b/test/registered/unit/multimodal/test_preprocess_cache.py index 2e26fbb59f..fcb9001a9f 100644 --- a/test/registered/unit/multimodal/test_preprocess_cache.py +++ b/test/registered/unit/multimodal/test_preprocess_cache.py @@ -25,7 +25,7 @@ from sglang.srt.runtime_context import publish, reset_context from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") class TestMediaIdentity(unittest.TestCase): diff --git a/test/registered/unit/multimodal/test_processor_async_call_sites.py b/test/registered/unit/multimodal/test_processor_async_call_sites.py index b1955fe722..398d43283b 100644 --- a/test/registered/unit/multimodal/test_processor_async_call_sites.py +++ b/test/registered/unit/multimodal/test_processor_async_call_sites.py @@ -18,7 +18,7 @@ import pytest from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=13, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") _MULTIMODAL_ROOT = ( pathlib.Path(__file__).resolve().parents[4] diff --git a/test/registered/unit/multimodal/test_processor_clone_isolation.py b/test/registered/unit/multimodal/test_processor_clone_isolation.py index 0ab717dfa6..6e1d2a8a0a 100644 --- a/test/registered/unit/multimodal/test_processor_clone_isolation.py +++ b/test/registered/unit/multimodal/test_processor_clone_isolation.py @@ -19,7 +19,7 @@ from sglang.srt.multimodal.processors.sarashina2_vision import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _NarrowImageProcessor: diff --git a/test/registered/unit/multimodal/test_processor_device_selection.py b/test/registered/unit/multimodal/test_processor_device_selection.py index 7d4cb89c51..019608b65b 100644 --- a/test/registered/unit/multimodal/test_processor_device_selection.py +++ b/test/registered/unit/multimodal/test_processor_device_selection.py @@ -17,7 +17,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") BASE = "sglang.srt.multimodal.processors.base_processor" diff --git a/test/registered/unit/multimodal/test_tensor_transport_mode.py b/test/registered/unit/multimodal/test_tensor_transport_mode.py index 18d550e542..2d459095e0 100644 --- a/test/registered/unit/multimodal/test_tensor_transport_mode.py +++ b/test/registered/unit/multimodal/test_tensor_transport_mode.py @@ -7,7 +7,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestTensorTransportMode(CustomTestCase): diff --git a/test/registered/unit/multimodal/test_vit_cuda_graph_runner.py b/test/registered/unit/multimodal/test_vit_cuda_graph_runner.py index c1c59de383..c8cd7a9c36 100644 --- a/test/registered/unit/multimodal/test_vit_cuda_graph_runner.py +++ b/test/registered/unit/multimodal/test_vit_cuda_graph_runner.py @@ -6,7 +6,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") from sglang.srt.multimodal.internvl_vit_cuda_graph_runner import ( InternViTCudaGraphRunner, diff --git a/test/registered/unit/multimodal/test_vit_npu_graph_runner.py b/test/registered/unit/multimodal/test_vit_npu_graph_runner.py index ebe2be3443..6d6c42dca2 100644 --- a/test/registered/unit/multimodal/test_vit_npu_graph_runner.py +++ b/test/registered/unit/multimodal/test_vit_npu_graph_runner.py @@ -8,7 +8,7 @@ import torch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") class _Block: diff --git a/test/registered/unit/npu/test_sparsity_driven_kv_offload_config.py b/test/registered/unit/npu/test_sparsity_driven_kv_offload_config.py index 8a136121f4..d6d144fb40 100644 --- a/test/registered/unit/npu/test_sparsity_driven_kv_offload_config.py +++ b/test/registered/unit/npu/test_sparsity_driven_kv_offload_config.py @@ -10,7 +10,7 @@ from sglang.srt.hardware_backend.npu.sparsity_driven_kv_offload.config import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_glm51_model_config(): diff --git a/test/registered/unit/observability/test_cpu_monitor.py b/test/registered/unit/observability/test_cpu_monitor.py index 3d0e18a58a..af0ec01276 100644 --- a/test/registered/unit/observability/test_cpu_monitor.py +++ b/test/registered/unit/observability/test_cpu_monitor.py @@ -6,7 +6,7 @@ from unittest.mock import MagicMock, patch from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=60, suite="base-a-test-cpu", nightly=True) +register_cpu_ci(est_time=9, suite="base-a-test-cpu", nightly=True) register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/observability/test_forward_pass_metrics.py b/test/registered/unit/observability/test_forward_pass_metrics.py index 1f8b1e171a..e7eee14cef 100644 --- a/test/registered/unit/observability/test_forward_pass_metrics.py +++ b/test/registered/unit/observability/test_forward_pass_metrics.py @@ -1,7 +1,7 @@ from sglang.srt.runtime_context import get_context, get_observability, get_parallel from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") import math import types diff --git a/test/registered/unit/observability/test_func_timer.py b/test/registered/unit/observability/test_func_timer.py index 6f646928cb..35be219cd8 100644 --- a/test/registered/unit/observability/test_func_timer.py +++ b/test/registered/unit/observability/test_func_timer.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") register_cpu_ci(est_time=4, suite="stage-b-test-cpu-intel") import asyncio diff --git a/test/registered/unit/observability/test_label_transform.py b/test/registered/unit/observability/test_label_transform.py index aa4857e03a..9fa6faa089 100644 --- a/test/registered/unit/observability/test_label_transform.py +++ b/test/registered/unit/observability/test_label_transform.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") register_cpu_ci(est_time=4, suite="stage-b-test-cpu-intel") import unittest diff --git a/test/registered/unit/observability/test_metrics_utils.py b/test/registered/unit/observability/test_metrics_utils.py index 11ed061830..1c21c63fe5 100644 --- a/test/registered/unit/observability/test_metrics_utils.py +++ b/test/registered/unit/observability/test_metrics_utils.py @@ -6,7 +6,7 @@ from sglang.srt.observability.utils import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/observability/test_ray_wrappers.py b/test/registered/unit/observability/test_ray_wrappers.py index 094be45297..5e9673d51d 100644 --- a/test/registered/unit/observability/test_ray_wrappers.py +++ b/test/registered/unit/observability/test_ray_wrappers.py @@ -25,7 +25,7 @@ from sglang.test.observability.fake_ray import ( ) from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # --------------------------------------------------------------------------- diff --git a/test/registered/unit/observability/test_req_time_stats.py b/test/registered/unit/observability/test_req_time_stats.py index bfb5ddf4f2..0a27a5961f 100644 --- a/test/registered/unit/observability/test_req_time_stats.py +++ b/test/registered/unit/observability/test_req_time_stats.py @@ -24,7 +24,7 @@ from sglang.srt.observability.trace import SpanAttributes from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestSetstatePreservesUnsetTimeSentinels(CustomTestCase): diff --git a/test/registered/unit/observability/test_request_metrics_exporter.py b/test/registered/unit/observability/test_request_metrics_exporter.py index 32f407c908..72200fb1bc 100644 --- a/test/registered/unit/observability/test_request_metrics_exporter.py +++ b/test/registered/unit/observability/test_request_metrics_exporter.py @@ -5,7 +5,7 @@ from typing import Any, Dict, List, Optional from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") import asyncio diff --git a/test/registered/unit/observability/test_scheduler_stage_metrics.py b/test/registered/unit/observability/test_scheduler_stage_metrics.py index b115acb956..5debbe39f5 100644 --- a/test/registered/unit/observability/test_scheduler_stage_metrics.py +++ b/test/registered/unit/observability/test_scheduler_stage_metrics.py @@ -15,7 +15,7 @@ from sglang.srt.observability.scheduler_stage_metrics import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestSchedulerStageMetricsRecorder(CustomTestCase): diff --git a/test/registered/unit/observability/test_startup_func_log_and_timer.py b/test/registered/unit/observability/test_startup_func_log_and_timer.py index e75497daf4..c445eafaa4 100644 --- a/test/registered/unit/observability/test_startup_func_log_and_timer.py +++ b/test/registered/unit/observability/test_startup_func_log_and_timer.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") register_cpu_ci(est_time=4, suite="stage-b-test-cpu-intel") import unittest diff --git a/test/registered/unit/observability/test_stat_loggers_di.py b/test/registered/unit/observability/test_stat_loggers_di.py index 9eb551fbcd..76dcba2bb9 100644 --- a/test/registered/unit/observability/test_stat_loggers_di.py +++ b/test/registered/unit/observability/test_stat_loggers_di.py @@ -20,7 +20,7 @@ GPU-backed metrics tests. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") import unittest from types import SimpleNamespace diff --git a/test/registered/unit/observability/test_trace.py b/test/registered/unit/observability/test_trace.py index 3c4cfffa17..3c872ecf7c 100644 --- a/test/registered/unit/observability/test_trace.py +++ b/test/registered/unit/observability/test_trace.py @@ -4,7 +4,7 @@ import os from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") import threading import unittest diff --git a/test/registered/unit/parser/test_code_completion_parser.py b/test/registered/unit/parser/test_code_completion_parser.py index 54262d123b..eae87be835 100644 --- a/test/registered/unit/parser/test_code_completion_parser.py +++ b/test/registered/unit/parser/test_code_completion_parser.py @@ -18,7 +18,7 @@ from sglang.srt.parser.code_completion_parser import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") register_cpu_ci(est_time=6, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/parser/test_conversation.py b/test/registered/unit/parser/test_conversation.py index 868bf4a865..1ec8afc519 100644 --- a/test/registered/unit/parser/test_conversation.py +++ b/test/registered/unit/parser/test_conversation.py @@ -32,7 +32,7 @@ from sglang.srt.parser.conversation import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/parser/test_harmony_parser.py b/test/registered/unit/parser/test_harmony_parser.py index e1e966b9c3..fd765247b8 100644 --- a/test/registered/unit/parser/test_harmony_parser.py +++ b/test/registered/unit/parser/test_harmony_parser.py @@ -12,7 +12,7 @@ from sglang.srt.parser.harmony_parser import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/parser/test_hunyuan_reasoning.py b/test/registered/unit/parser/test_hunyuan_reasoning.py index d550fcae5c..36ec4b315d 100644 --- a/test/registered/unit/parser/test_hunyuan_reasoning.py +++ b/test/registered/unit/parser/test_hunyuan_reasoning.py @@ -7,7 +7,7 @@ from sglang.srt.parser.hunyuan_reasoning import normalize_hunyuan_reasoning_effo from sglang.srt.parser.template_detection import ReasoningToggleConfig from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") @pytest.mark.parametrize( diff --git a/test/registered/unit/parser/test_inkling_renderer.py b/test/registered/unit/parser/test_inkling_renderer.py index 2fad76f0e7..a3d788bc94 100644 --- a/test/registered/unit/parser/test_inkling_renderer.py +++ b/test/registered/unit/parser/test_inkling_renderer.py @@ -19,7 +19,7 @@ from sglang.srt.parser.inkling_tokenizer import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _text(value: str) -> list[int]: diff --git a/test/registered/unit/parser/test_jinja_template_utils.py b/test/registered/unit/parser/test_jinja_template_utils.py index 53dc78a616..716e356cc6 100644 --- a/test/registered/unit/parser/test_jinja_template_utils.py +++ b/test/registered/unit/parser/test_jinja_template_utils.py @@ -11,7 +11,7 @@ from sglang.srt.utils import VideoData from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") register_cpu_ci(est_time=6, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/parser/test_k2_v3_reasoning_parser.py b/test/registered/unit/parser/test_k2_v3_reasoning_parser.py index 2169dbfdb4..9d107e4a85 100644 --- a/test/registered/unit/parser/test_k2_v3_reasoning_parser.py +++ b/test/registered/unit/parser/test_k2_v3_reasoning_parser.py @@ -8,7 +8,7 @@ from sglang.srt.parser.reasoning_parser import K2V3Detector, ReasoningParser from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class TestK2V3ReasoningParser(CustomTestCase): diff --git a/test/registered/unit/parser/test_kimik3_reasoning_parser.py b/test/registered/unit/parser/test_kimik3_reasoning_parser.py index b9fda71655..4fa0524601 100644 --- a/test/registered/unit/parser/test_kimik3_reasoning_parser.py +++ b/test/registered/unit/parser/test_kimik3_reasoning_parser.py @@ -14,7 +14,7 @@ from sglang.srt.function_call.kimik3_format import ( from sglang.srt.parser.reasoning_parser import KimiK3Detector, ReasoningParser from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") def _stream(detector: KimiK3Detector, chunks: list[str]) -> tuple[str, str]: diff --git a/test/registered/unit/parser/test_reasoning_parser.py b/test/registered/unit/parser/test_reasoning_parser.py index acf2152d74..39e2aa4f06 100644 --- a/test/registered/unit/parser/test_reasoning_parser.py +++ b/test/registered/unit/parser/test_reasoning_parser.py @@ -24,7 +24,7 @@ from sglang.srt.parser.reasoning_parser_names import REASONING_PARSER_NAMES from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestBaseReasoningFormatDetector(CustomTestCase): diff --git a/test/registered/unit/platforms/test_platform_interface.py b/test/registered/unit/platforms/test_platform_interface.py index 1e36b9b437..f6bb8dea61 100644 --- a/test/registered/unit/platforms/test_platform_interface.py +++ b/test/registered/unit/platforms/test_platform_interface.py @@ -25,7 +25,7 @@ from sglang.srt.platforms.xpu import XpuSRTPlatform from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") # --------------------------------------------------------------------------- diff --git a/test/registered/unit/plugins/test_hook_registry.py b/test/registered/unit/plugins/test_hook_registry.py index 41eb4fb3a5..de251bb90c 100644 --- a/test/registered/unit/plugins/test_hook_registry.py +++ b/test/registered/unit/plugins/test_hook_registry.py @@ -16,7 +16,7 @@ from sglang.srt.plugins.hook_registry import HookRegistry, HookType, plugin_hook from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") # --------------------------------------------------------------------------- # Helpers: synthetic module creation diff --git a/test/registered/unit/plugins/test_load_plugins.py b/test/registered/unit/plugins/test_load_plugins.py index 3b84f88934..48df22428f 100644 --- a/test/registered/unit/plugins/test_load_plugins.py +++ b/test/registered/unit/plugins/test_load_plugins.py @@ -19,7 +19,7 @@ from sglang.srt.plugins import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _make_ep(name, dist_name=None, load_fn=None): diff --git a/test/registered/unit/sampling/test_ascend_sampler.py b/test/registered/unit/sampling/test_ascend_sampler.py index 70786258a8..fa75040dc9 100644 --- a/test/registered/unit/sampling/test_ascend_sampler.py +++ b/test/registered/unit/sampling/test_ascend_sampler.py @@ -9,7 +9,7 @@ import torch from sglang.srt.layers import sampler as sampler_module from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=1, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestAscendSamplerDispatch(unittest.TestCase): diff --git a/test/registered/unit/sampling/test_custom_logit_processor.py b/test/registered/unit/sampling/test_custom_logit_processor.py index dc3b1b42bd..57f3310a9a 100644 --- a/test/registered/unit/sampling/test_custom_logit_processor.py +++ b/test/registered/unit/sampling/test_custom_logit_processor.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") import json diff --git a/test/registered/unit/sampling/test_penaltylib.py b/test/registered/unit/sampling/test_penaltylib.py index 3638c62c95..c1db5ec6e0 100644 --- a/test/registered/unit/sampling/test_penaltylib.py +++ b/test/registered/unit/sampling/test_penaltylib.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") import unittest diff --git a/test/registered/unit/sampling/test_sampling_batch_info.py b/test/registered/unit/sampling/test_sampling_batch_info.py index 016f6d2341..a05beae34e 100644 --- a/test/registered/unit/sampling/test_sampling_batch_info.py +++ b/test/registered/unit/sampling/test_sampling_batch_info.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=24, suite="base-a-test-cpu") +register_cpu_ci(est_time=15, suite="base-a-test-cpu") register_cpu_ci(est_time=6, suite="stage-b-test-cpu-intel") import unittest diff --git a/test/registered/unit/sampling/test_sampling_metadata_staging.py b/test/registered/unit/sampling/test_sampling_metadata_staging.py index f469ae0c36..33eb800b3d 100644 --- a/test/registered/unit/sampling/test_sampling_metadata_staging.py +++ b/test/registered/unit/sampling/test_sampling_metadata_staging.py @@ -16,7 +16,7 @@ from sglang.srt.sampling.sampling_params import SamplingParams from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=15, suite="base-a-test-cpu") +register_cpu_ci(est_time=11, suite="base-a-test-cpu") register_cuda_ci(est_time=20, stage="base-b-kernel-unit", runner_config="1-gpu-large") VOCAB_SIZE = 32 diff --git a/test/registered/unit/sampling/test_sampling_params.py b/test/registered/unit/sampling/test_sampling_params.py index 7edb6f8604..526fad96a9 100644 --- a/test/registered/unit/sampling/test_sampling_params.py +++ b/test/registered/unit/sampling/test_sampling_params.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci, register_xpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") register_xpu_ci(est_time=10, suite="stage-a-test-1-gpu-xpu") diff --git a/test/registered/unit/server_args/test_declared_fallbacks.py b/test/registered/unit/server_args/test_declared_fallbacks.py index f0601c0956..1e1096e494 100644 --- a/test/registered/unit/server_args/test_declared_fallbacks.py +++ b/test/registered/unit/server_args/test_declared_fallbacks.py @@ -40,7 +40,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=30, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") def _resolved(**kwargs) -> ServerArgs: diff --git a/test/registered/unit/server_args/test_model_config_cache.py b/test/registered/unit/server_args/test_model_config_cache.py index 0337e68859..342f54a7a3 100644 --- a/test/registered/unit/server_args/test_model_config_cache.py +++ b/test/registered/unit/server_args/test_model_config_cache.py @@ -28,7 +28,7 @@ from sglang.srt.utils import runai_utils from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") _MINI_CONFIG = { "architectures": ["LlamaForCausalLM"], diff --git a/test/registered/unit/server_args/test_model_source_paths.py b/test/registered/unit/server_args/test_model_source_paths.py index ef128727e5..cbe06fea24 100644 --- a/test/registered/unit/server_args/test_model_source_paths.py +++ b/test/registered/unit/server_args/test_model_source_paths.py @@ -24,7 +24,7 @@ the model-configuration cache notices. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") import json import os diff --git a/test/registered/unit/server_args/test_page_major_backend_allowlist.py b/test/registered/unit/server_args/test_page_major_backend_allowlist.py index 5cb295b144..f4e41bfa21 100644 --- a/test/registered/unit/server_args/test_page_major_backend_allowlist.py +++ b/test/registered/unit/server_args/test_page_major_backend_allowlist.py @@ -38,7 +38,7 @@ from sglang.srt.configs.model_config import AttentionArch from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _accepts( diff --git a/test/registered/unit/server_args/test_platform_prefill_cp_deprecation.py b/test/registered/unit/server_args/test_platform_prefill_cp_deprecation.py index bd3c23bfd2..6fa0bb169b 100644 --- a/test/registered/unit/server_args/test_platform_prefill_cp_deprecation.py +++ b/test/registered/unit/server_args/test_platform_prefill_cp_deprecation.py @@ -11,7 +11,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestPlatformPrefillCPDeprecation(CustomTestCase): diff --git a/test/registered/unit/server_args/test_record_holds_the_raw_input.py b/test/registered/unit/server_args/test_record_holds_the_raw_input.py index f9b153511d..8f557c8e1b 100644 --- a/test/registered/unit/server_args/test_record_holds_the_raw_input.py +++ b/test/registered/unit/server_args/test_record_holds_the_raw_input.py @@ -21,7 +21,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") _MINI_CONFIG = { diff --git a/test/registered/unit/server_args/test_resolution_declarations.py b/test/registered/unit/server_args/test_resolution_declarations.py index 9fe1f63e1c..ea44593673 100644 --- a/test/registered/unit/server_args/test_resolution_declarations.py +++ b/test/registered/unit/server_args/test_resolution_declarations.py @@ -25,7 +25,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=26, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") _SRT = pathlib.Path(sglang.__file__).resolve().parent / "srt" diff --git a/test/registered/unit/server_args/test_resolution_hook_registry.py b/test/registered/unit/server_args/test_resolution_hook_registry.py index 670d106c24..ff63954af5 100644 --- a/test/registered/unit/server_args/test_resolution_hook_registry.py +++ b/test/registered/unit/server_args/test_resolution_hook_registry.py @@ -4,7 +4,7 @@ end-to-end proof that an override reaches a real `resolve_once()`.""" from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") import ast import glob diff --git a/test/registered/unit/server_args/test_resolution_is_reproducible.py b/test/registered/unit/server_args/test_resolution_is_reproducible.py index 4ed946b1ce..544b925e2d 100644 --- a/test/registered/unit/server_args/test_resolution_is_reproducible.py +++ b/test/registered/unit/server_args/test_resolution_is_reproducible.py @@ -51,13 +51,13 @@ from sglang.test.ci.ci_register import ( ) from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=104, suite="base-a-test-cpu") +register_cpu_ci(est_time=11, suite="base-a-test-cpu") # Also on a GPU runner: the resolution branches that matter most (backend # defaults, DeepSeek handlers, capability gates) go through `is_cuda()` / # `is_hip()` / device capability, which inspect the actual hardware -- passing # device="cuda" on a CPU box does not reach them, so a leak confined to a GPU # handler would never fail the CPU registration alone. -register_cuda_ci(est_time=78, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=22, stage="base-b", runner_config="1-gpu-small") # ROCm too: `is_hip()` gates its own set of backend and DeepSeek handlers, which # neither the CPU suite nor a CUDA runner reaches. register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd") diff --git a/test/registered/unit/server_args/test_server_args.py b/test/registered/unit/server_args/test_server_args.py index fed3493e58..ec20b298c6 100644 --- a/test/registered/unit/server_args/test_server_args.py +++ b/test/registered/unit/server_args/test_server_args.py @@ -109,7 +109,7 @@ from sglang.test.test_utils import ( CustomTestCase, ) -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") register_cpu_ci(est_time=11, suite="stage-b-test-cpu-intel") # Mock get_device() so all tests run on CPU-only CI runners diff --git a/test/registered/unit/server_args/test_unified_prefill_cuda_graph_gate.py b/test/registered/unit/server_args/test_unified_prefill_cuda_graph_gate.py index ac79697f07..84011be7d0 100644 --- a/test/registered/unit/server_args/test_unified_prefill_cuda_graph_gate.py +++ b/test/registered/unit/server_args/test_unified_prefill_cuda_graph_gate.py @@ -44,7 +44,7 @@ from sglang.srt.model_executor.cuda_graph_config import Backend from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") def _run_handler(*, prefill_backend, attention_backends): diff --git a/test/registered/unit/server_args/test_unified_tbo_gate.py b/test/registered/unit/server_args/test_unified_tbo_gate.py index f0f62c7c55..750a89b07a 100644 --- a/test/registered/unit/server_args/test_unified_tbo_gate.py +++ b/test/registered/unit/server_args/test_unified_tbo_gate.py @@ -30,7 +30,7 @@ from sglang.srt.model_executor.cuda_graph_config import Backend from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") def _run_handler(*, unified, tbo): diff --git a/test/registered/unit/spec/test_adaptive_runtime_state.py b/test/registered/unit/spec/test_adaptive_runtime_state.py index 2cc15d89e9..a342d050d5 100644 --- a/test/registered/unit/spec/test_adaptive_runtime_state.py +++ b/test/registered/unit/spec/test_adaptive_runtime_state.py @@ -6,7 +6,7 @@ from sglang.srt.speculative.adaptive_runtime_state import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=4, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_state(steps: int) -> SpecRuntimeState: diff --git a/test/registered/unit/spec/test_adaptive_spec_params.py b/test/registered/unit/spec/test_adaptive_spec_params.py index f9bcf80e88..83195c53b5 100644 --- a/test/registered/unit/spec/test_adaptive_spec_params.py +++ b/test/registered/unit/spec/test_adaptive_spec_params.py @@ -10,7 +10,7 @@ from sglang.srt.speculative.adaptive_spec_params import ( from sglang.test.ci.ci_register import register_cpu_ci, register_xpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") register_xpu_ci(est_time=10, suite="stage-a-test-1-gpu-xpu") diff --git a/test/registered/unit/spec/test_decode_bookkeeping_ownership.py b/test/registered/unit/spec/test_decode_bookkeeping_ownership.py index cea26addd3..4fa7556f10 100644 --- a/test/registered/unit/spec/test_decode_bookkeeping_ownership.py +++ b/test/registered/unit/spec/test_decode_bookkeeping_ownership.py @@ -20,7 +20,7 @@ from pathlib import Path from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=22, suite="base-a-test-cpu") +register_cpu_ci(est_time=18, suite="base-a-test-cpu") _REPO_ROOT = Path(__file__).resolve().parents[4] _SRT_DIR = _REPO_ROOT / "python" / "sglang" / "srt" diff --git a/test/registered/unit/spec/test_dflash_extra_buffer_lazy.py b/test/registered/unit/spec/test_dflash_extra_buffer_lazy.py index fa2a3dca72..7b579bd3d7 100644 --- a/test/registered/unit/spec/test_dflash_extra_buffer_lazy.py +++ b/test/registered/unit/spec/test_dflash_extra_buffer_lazy.py @@ -13,7 +13,7 @@ from sglang.srt.arg_groups.mamba_hook import validate_mamba_extra_buffer from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small") from sglang.srt.model_executor.forward_batch_info import ForwardMode from sglang.srt.runtime_context import override_platform diff --git a/test/registered/unit/spec/test_dflash_logits.py b/test/registered/unit/spec/test_dflash_logits.py index f635a7b7f2..c58f34c6b7 100644 --- a/test/registered/unit/spec/test_dflash_logits.py +++ b/test/registered/unit/spec/test_dflash_logits.py @@ -12,7 +12,7 @@ from sglang.srt.models.dflash import ( from sglang.srt.speculative.dflash_utils import parse_dflash_draft_config from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=38, suite="base-a-test-cpu") +register_cpu_ci(est_time=25, suite="base-a-test-cpu") def test_dflash_unary_logit_transform(): diff --git a/test/registered/unit/spec/test_dflash_overlap_hostsync.py b/test/registered/unit/spec/test_dflash_overlap_hostsync.py index 809f3466a7..e6a20d7419 100644 --- a/test/registered/unit/spec/test_dflash_overlap_hostsync.py +++ b/test/registered/unit/spec/test_dflash_overlap_hostsync.py @@ -12,7 +12,7 @@ from sglang.srt.runtime_context import get_context from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=23, stage="base-b", runner_config="1-gpu-small") _HAS_CUDA = torch.cuda.is_available() diff --git a/test/registered/unit/spec/test_draft_construction_isolation.py b/test/registered/unit/spec/test_draft_construction_isolation.py index bf0aedd93e..e1ab15f88c 100644 --- a/test/registered/unit/spec/test_draft_construction_isolation.py +++ b/test/registered/unit/spec/test_draft_construction_isolation.py @@ -23,7 +23,7 @@ from sglang.srt.runtime_context import get_context, get_flags, get_model, publis from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") class _AlwaysDisables: diff --git a/test/registered/unit/spec/test_draft_per_runner_config.py b/test/registered/unit/spec/test_draft_per_runner_config.py index c189236171..c5232b036f 100644 --- a/test/registered/unit/spec/test_draft_per_runner_config.py +++ b/test/registered/unit/spec/test_draft_per_runner_config.py @@ -26,7 +26,7 @@ from sglang.srt.runtime_context import get_context, get_model from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") class _StopConstruction(Exception): diff --git a/test/registered/unit/spec/test_dspark_target_hidden_projection.py b/test/registered/unit/spec/test_dspark_target_hidden_projection.py index 2f85aa8f0b..d3083b9e70 100644 --- a/test/registered/unit/spec/test_dspark_target_hidden_projection.py +++ b/test/registered/unit/spec/test_dspark_target_hidden_projection.py @@ -14,7 +14,7 @@ from sglang.srt.speculative.dspark_components.dspark_worker_v2 import DSparkWork from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=11, suite="base-a-test-cpu") class _Attention: diff --git a/test/registered/unit/spec/test_eagle_draft_cuda_graph_runner.py b/test/registered/unit/spec/test_eagle_draft_cuda_graph_runner.py index 25dfe2adcc..a1a2d6da26 100644 --- a/test/registered/unit/spec/test_eagle_draft_cuda_graph_runner.py +++ b/test/registered/unit/spec/test_eagle_draft_cuda_graph_runner.py @@ -27,7 +27,7 @@ from sglang.srt.speculative.eagle_draft_cuda_graph_runner import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") CAPTURE_BS = 4 SEQ_LEN_FILL_VALUE = 1 diff --git a/test/registered/unit/spec/test_eagle_draft_extend_logits.py b/test/registered/unit/spec/test_eagle_draft_extend_logits.py index f02c6970c0..55478b9489 100644 --- a/test/registered/unit/spec/test_eagle_draft_extend_logits.py +++ b/test/registered/unit/spec/test_eagle_draft_extend_logits.py @@ -10,7 +10,7 @@ from sglang.srt.model_executor.forward_batch_info import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestEagleDraftExtendLogitsPruning(unittest.TestCase): diff --git a/test/registered/unit/spec/test_eagle_gate_routing.py b/test/registered/unit/spec/test_eagle_gate_routing.py index 145206c777..d956749070 100644 --- a/test/registered/unit/spec/test_eagle_gate_routing.py +++ b/test/registered/unit/spec/test_eagle_gate_routing.py @@ -19,7 +19,7 @@ from sglang.srt.speculative.eagle_utils import _verify_uses_greedy from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") def _pre_patch_gate(is_all_greedy, is_cpu, is_hip, is_xpu): diff --git a/test/registered/unit/spec/test_eagle_seeded_coins.py b/test/registered/unit/spec/test_eagle_seeded_coins.py index 69905f237c..3e7c7cf50f 100644 --- a/test/registered/unit/spec/test_eagle_seeded_coins.py +++ b/test/registered/unit/spec/test_eagle_seeded_coins.py @@ -21,7 +21,7 @@ from sglang.srt.speculative.eagle_utils import _seeded_verify_coins, _verify_coi from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small") DRAFT_TOKEN_NUM = 4 diff --git a/test/registered/unit/spec/test_eagle_worker_v2_topk1_fastpath.py b/test/registered/unit/spec/test_eagle_worker_v2_topk1_fastpath.py index 3f21212f12..5dfe03cbe2 100644 --- a/test/registered/unit/spec/test_eagle_worker_v2_topk1_fastpath.py +++ b/test/registered/unit/spec/test_eagle_worker_v2_topk1_fastpath.py @@ -25,7 +25,7 @@ from sglang.test.test_utils import CustomTestCase register_amd_ci(est_time=20, stage="stage-b", runner_config="1-gpu-small-amd") -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") DEVICE = "cuda" if torch.cuda.is_available() else "cpu" diff --git a/test/registered/unit/spec/test_fast_prefill_plan.py b/test/registered/unit/spec/test_fast_prefill_plan.py index 6e772b30b9..809acd1c32 100644 --- a/test/registered/unit/spec/test_fast_prefill_plan.py +++ b/test/registered/unit/spec/test_fast_prefill_plan.py @@ -27,7 +27,7 @@ try: except ImportError: _HAS_FLASHINFER = False -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") # Draft-extend layout: constant qo (num_tokens_per_req per req), page_size 1. # Non-uniform seq_lens make cumsum non-trivial, so a wrong per-row kv split is diff --git a/test/registered/unit/spec/test_ngram_corpus.py b/test/registered/unit/spec/test_ngram_corpus.py index 5c0c45a056..adcfc3ba0c 100644 --- a/test/registered/unit/spec/test_ngram_corpus.py +++ b/test/registered/unit/spec/test_ngram_corpus.py @@ -13,7 +13,7 @@ from sglang.srt.speculative.cpp_ngram.ngram_corpus import NgramCorpus from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=20, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") def _make_corpus(match_type="BFS", **kwargs): diff --git a/test/registered/unit/spec/test_ngram_mamba_verify_update.py b/test/registered/unit/spec/test_ngram_mamba_verify_update.py index fe062eb256..5b0a319ba2 100644 --- a/test/registered/unit/spec/test_ngram_mamba_verify_update.py +++ b/test/registered/unit/spec/test_ngram_mamba_verify_update.py @@ -7,7 +7,7 @@ import sglang.srt from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=23, suite="base-a-test-cpu") +register_cpu_ci(est_time=15, suite="base-a-test-cpu") class TestNgramLastCorrectStepIndices(CustomTestCase): diff --git a/test/registered/unit/spec/test_oot_dflash_hooks.py b/test/registered/unit/spec/test_oot_dflash_hooks.py index 6e5bcc5405..f482407416 100644 --- a/test/registered/unit/spec/test_oot_dflash_hooks.py +++ b/test/registered/unit/spec/test_oot_dflash_hooks.py @@ -11,7 +11,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=2, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") HOOK_MODULE = "sglang.srt.arg_groups.speculative_hook" diff --git a/test/registered/unit/spec/test_plugin_hook_signatures.py b/test/registered/unit/spec/test_plugin_hook_signatures.py index 65e0b4b847..8732e6d74a 100644 --- a/test/registered/unit/spec/test_plugin_hook_signatures.py +++ b/test/registered/unit/spec/test_plugin_hook_signatures.py @@ -25,7 +25,7 @@ from sglang.srt.speculative.spec_registry import CustomSpecAlgo from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") def _dispatched_methods(): diff --git a/test/registered/unit/spec/test_spec_cpu_overlap_constraint.py b/test/registered/unit/spec/test_spec_cpu_overlap_constraint.py index 88cf080c88..5c17f17f09 100644 --- a/test/registered/unit/spec/test_spec_cpu_overlap_constraint.py +++ b/test/registered/unit/spec/test_spec_cpu_overlap_constraint.py @@ -7,7 +7,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") def _make_spec_args(device: str, algorithm: str = "EAGLE", **overrides) -> ServerArgs: diff --git a/test/registered/unit/spec/test_spec_utils_traverse_tree.py b/test/registered/unit/spec/test_spec_utils_traverse_tree.py index 5061d0d43e..e25a302c22 100644 --- a/test/registered/unit/spec/test_spec_utils_traverse_tree.py +++ b/test/registered/unit/spec/test_spec_utils_traverse_tree.py @@ -14,7 +14,7 @@ import torch from sglang.srt.speculative.spec_utils import GrammarTree, traverse_tree from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestTraverseTreePassesIntsToGrammar(unittest.TestCase): diff --git a/test/registered/unit/spec/test_uno_tree_config.py b/test/registered/unit/spec/test_uno_tree_config.py index 76cf4c4b2b..7f58896c4a 100644 --- a/test/registered/unit/spec/test_uno_tree_config.py +++ b/test/registered/unit/spec/test_uno_tree_config.py @@ -8,7 +8,7 @@ from sglang.srt.arg_groups.speculative_hook import _handle_uno from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=11, suite="base-a-test-cpu") class TestUnoTreeConfig(CustomTestCase): diff --git a/test/registered/unit/spec/test_uno_tree_sparse_sampling.py b/test/registered/unit/spec/test_uno_tree_sparse_sampling.py index e5c5314f6f..fd0a44c865 100644 --- a/test/registered/unit/spec/test_uno_tree_sparse_sampling.py +++ b/test/registered/unit/spec/test_uno_tree_sparse_sampling.py @@ -13,7 +13,7 @@ from sglang.srt.speculative.uno_utils import sample_uno_tree_target_tokens from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=8, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestUnoTreeSparseSampling(CustomTestCase): diff --git a/test/registered/unit/state_capturer/test_routed_experts_scattered_a2a.py b/test/registered/unit/state_capturer/test_routed_experts_scattered_a2a.py index c8e3654dfb..7a008f27fb 100644 --- a/test/registered/unit/state_capturer/test_routed_experts_scattered_a2a.py +++ b/test/registered/unit/state_capturer/test_routed_experts_scattered_a2a.py @@ -11,7 +11,7 @@ from sglang.srt.state_capturer import routed_experts as re_mod from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=12, suite="base-a-test-cpu") class TestScatteredA2ABackendHelper(CustomTestCase): diff --git a/test/registered/unit/test_checkpoint_quantization.py b/test/registered/unit/test_checkpoint_quantization.py index 5f6f1a4734..de2b145fc8 100644 --- a/test/registered/unit/test_checkpoint_quantization.py +++ b/test/registered/unit/test_checkpoint_quantization.py @@ -13,7 +13,7 @@ from sglang.srt.model_loader.checkpoint_quantization import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestResolveCheckpointQuantSpec(CustomTestCase): diff --git a/test/registered/unit/test_cuda_vmm_utils.py b/test/registered/unit/test_cuda_vmm_utils.py index 374432a715..ffa20d5d8b 100644 --- a/test/registered/unit/test_cuda_vmm_utils.py +++ b/test/registered/unit/test_cuda_vmm_utils.py @@ -37,7 +37,7 @@ from sglang.srt.utils.cuda_vmm_utils import ( from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.kernels.utils import multigpu_pytest_main -register_cuda_ci(est_time=15, stage="base-b", runner_config="2-gpu-large") +register_cuda_ci(est_time=11, stage="base-b", runner_config="2-gpu-large") _FABRIC = drv.CUmemAllocationHandleType.CU_MEM_HANDLE_TYPE_FABRIC _POSIX_FD = drv.CUmemAllocationHandleType.CU_MEM_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR diff --git a/test/registered/unit/test_dsa_tilelang_fp8_validation.py b/test/registered/unit/test_dsa_tilelang_fp8_validation.py index 42ed533fc7..b10b315eb4 100644 --- a/test/registered/unit/test_dsa_tilelang_fp8_validation.py +++ b/test/registered/unit/test_dsa_tilelang_fp8_validation.py @@ -11,7 +11,7 @@ from sglang.srt.arg_groups.overrides import _check_tilelang_dsa_fp8_kv from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") class TestDsaTilelangFp8Validation(CustomTestCase): diff --git a/test/registered/unit/test_environ.py b/test/registered/unit/test_environ.py index f75a37a9aa..6352119baa 100644 --- a/test/registered/unit/test_environ.py +++ b/test/registered/unit/test_environ.py @@ -11,7 +11,7 @@ from contextlib import ExitStack from sglang.srt.environ import _DEPRECATED_ENVS, _DeprecatedEnv, envs from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestEnvField(unittest.TestCase): diff --git a/test/registered/unit/test_flashinfer_sparse_mla.py b/test/registered/unit/test_flashinfer_sparse_mla.py index 1b72d94403..f6d231b0a5 100644 --- a/test/registered/unit/test_flashinfer_sparse_mla.py +++ b/test/registered/unit/test_flashinfer_sparse_mla.py @@ -11,7 +11,7 @@ from sglang.kernels.ops.attention.flash_mla_sm120 import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") class TestFlashInferSparseMLAAdapter(unittest.TestCase): diff --git a/test/registered/unit/test_global_config_read_ratchet.py b/test/registered/unit/test_global_config_read_ratchet.py index a623462bff..7e745f9f3a 100644 --- a/test/registered/unit/test_global_config_read_ratchet.py +++ b/test/registered/unit/test_global_config_read_ratchet.py @@ -27,7 +27,7 @@ caller decided which instance to hand over. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=16, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") import ast import unittest diff --git a/test/registered/unit/test_model_overrides.py b/test/registered/unit/test_model_overrides.py index 594096c169..1431b9a795 100644 --- a/test/registered/unit/test_model_overrides.py +++ b/test/registered/unit/test_model_overrides.py @@ -3,7 +3,7 @@ gate, publish wiring, and the per-arch golden diffs for migrated families.""" from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=5, suite="base-a-test-cpu") import dataclasses import json diff --git a/test/registered/unit/test_platform_context.py b/test/registered/unit/test_platform_context.py index 201c2a438e..4db48f5183 100644 --- a/test/registered/unit/test_platform_context.py +++ b/test/registered/unit/test_platform_context.py @@ -16,7 +16,7 @@ from sglang.srt.utils import common as _common from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") class TestPlatformContext(CustomTestCase): diff --git a/test/registered/unit/test_pre_publish_readers.py b/test/registered/unit/test_pre_publish_readers.py index 4fb7c464d3..253a2c2313 100644 --- a/test/registered/unit/test_pre_publish_readers.py +++ b/test/registered/unit/test_pre_publish_readers.py @@ -15,7 +15,7 @@ requiring a full server boot. from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=4, suite="base-a-test-cpu") import logging import unittest diff --git a/test/registered/unit/test_quantization_post_load.py b/test/registered/unit/test_quantization_post_load.py index 11cf2e6bde..028a4e90b4 100644 --- a/test/registered/unit/test_quantization_post_load.py +++ b/test/registered/unit/test_quantization_post_load.py @@ -9,7 +9,7 @@ from sglang.srt.model_loader.loader import device_loading_context from sglang.srt.model_loader.post_load import stage_module_for_post_load from sglang.test.ci.ci_register import register_cuda_ci -register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small") def _process_device() -> torch.device | None: diff --git a/test/registered/unit/test_rocm_pageable_h2d_staging.py b/test/registered/unit/test_rocm_pageable_h2d_staging.py index d8693df01f..000210f8f3 100644 --- a/test/registered/unit/test_rocm_pageable_h2d_staging.py +++ b/test/registered/unit/test_rocm_pageable_h2d_staging.py @@ -13,7 +13,7 @@ from sglang.srt.arg_groups.platform_hook import handle_amd_specifics from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=5, suite="base-a-test-cpu") +register_cpu_ci(est_time=9, suite="base-a-test-cpu") VAR = "GPU_PINNED_MIN_XFER_SIZE" diff --git a/test/registered/unit/test_runai_utils.py b/test/registered/unit/test_runai_utils.py index 56c50b37e9..92c67878d7 100644 --- a/test/registered/unit/test_runai_utils.py +++ b/test/registered/unit/test_runai_utils.py @@ -5,7 +5,7 @@ from sglang.srt.utils.runai_utils import ObjectStorageModel, is_runai_obj_uri from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") register_cpu_ci(est_time=6, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/test_runtime_context.py b/test/registered/unit/test_runtime_context.py index b7bd5faa9b..432cf5cae6 100644 --- a/test/registered/unit/test_runtime_context.py +++ b/test/registered/unit/test_runtime_context.py @@ -2,7 +2,7 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=33, suite="base-a-test-cpu") +register_cpu_ci(est_time=46, suite="base-a-test-cpu") import dataclasses import json diff --git a/test/registered/unit/test_runtime_context_config_bags.py b/test/registered/unit/test_runtime_context_config_bags.py index 27a9ed635d..6b843246c6 100644 --- a/test/registered/unit/test_runtime_context_config_bags.py +++ b/test/registered/unit/test_runtime_context_config_bags.py @@ -19,7 +19,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") @dataclasses.dataclass diff --git a/test/registered/unit/test_runtime_context_override.py b/test/registered/unit/test_runtime_context_override.py index 25185adf93..8238848884 100644 --- a/test/registered/unit/test_runtime_context_override.py +++ b/test/registered/unit/test_runtime_context_override.py @@ -15,7 +15,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestContextOverride(CustomTestCase): diff --git a/test/registered/unit/test_server_args_cli_metadata.py b/test/registered/unit/test_server_args_cli_metadata.py index 4ebf060d6c..dd68d60fb4 100644 --- a/test/registered/unit/test_server_args_cli_metadata.py +++ b/test/registered/unit/test_server_args_cli_metadata.py @@ -8,7 +8,7 @@ from sglang.srt.utils.common import human_readable_int from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=7, suite="base-a-test-cpu") class TestServerArgsMigratedCliMetadata(CustomTestCase): diff --git a/test/registered/unit/test_server_args_namespaces.py b/test/registered/unit/test_server_args_namespaces.py index 6767ec8efe..b422dbae8b 100644 --- a/test/registered/unit/test_server_args_namespaces.py +++ b/test/registered/unit/test_server_args_namespaces.py @@ -21,7 +21,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=31, suite="base-a-test-cpu") +register_cpu_ci(est_time=20, suite="base-a-test-cpu") # Locked taxonomy (global_context/11-server-args-namespace-split.md). VALID_NAMESPACES = { diff --git a/test/registered/unit/test_split_attention_backend_decisions.py b/test/registered/unit/test_split_attention_backend_decisions.py index 5d73b2dadb..93c2dbc7f5 100644 --- a/test/registered/unit/test_split_attention_backend_decisions.py +++ b/test/registered/unit/test_split_attention_backend_decisions.py @@ -27,7 +27,7 @@ from sglang.srt.server_args import ServerArgs from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=10, suite="base-a-test-cpu") from sglang.srt.arg_groups.overrides import attention_backends_of, resolved_view diff --git a/test/registered/unit/tokenizer/test_mistral_empty_assistant.py b/test/registered/unit/tokenizer/test_mistral_empty_assistant.py index 287d24fc4f..95f7e75cbe 100644 --- a/test/registered/unit/tokenizer/test_mistral_empty_assistant.py +++ b/test/registered/unit/tokenizer/test_mistral_empty_assistant.py @@ -5,7 +5,7 @@ from sglang.srt.utils.hf_transformers.mistral_utils import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") def _user(text): diff --git a/test/registered/unit/tokenizer/test_tekken_tokenizer_routing.py b/test/registered/unit/tokenizer/test_tekken_tokenizer_routing.py index 8a96bfb54e..eaa49baf29 100644 --- a/test/registered/unit/tokenizer/test_tekken_tokenizer_routing.py +++ b/test/registered/unit/tokenizer/test_tekken_tokenizer_routing.py @@ -8,7 +8,7 @@ import unittest from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=3, suite="base-a-test-cpu") from sglang.srt.utils.hf_transformers.mistral_utils import is_bare_tekken_checkpoint from sglang.srt.utils.hf_transformers.tokenizer import get_tokenizer diff --git a/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py b/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py index 115d6dd5cb..2077be9cf6 100644 --- a/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py +++ b/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py @@ -6,7 +6,7 @@ from unittest.mock import MagicMock, patch from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") from sglang.srt.tokenizer.tiktoken_tokenizer import ( CONTROL_TOKEN_TEXTS, diff --git a/test/registered/unit/utils/test_common.py b/test/registered/unit/utils/test_common.py index 2b28b089d6..73324b5f0c 100644 --- a/test/registered/unit/utils/test_common.py +++ b/test/registered/unit/utils/test_common.py @@ -14,7 +14,7 @@ from sglang.srt.utils.common import ( from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=5, stage="stage-b", runner_config="1-gpu-small-amd") diff --git a/test/registered/unit/utils/test_diffusion_torch_fallback.py b/test/registered/unit/utils/test_diffusion_torch_fallback.py index f4325b3fdb..eecf55106d 100644 --- a/test/registered/unit/utils/test_diffusion_torch_fallback.py +++ b/test/registered/unit/utils/test_diffusion_torch_fallback.py @@ -13,7 +13,7 @@ from sglang.kernels.ops.diffusion.common.fallback_torch import ( ) from sglang.test.ci.ci_register import register_cpu_ci, register_mlx_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=1, suite="stage-a-test-cpu-intel") register_cpu_ci(est_time=1, suite="base-b-test-cpu-arm64") register_mlx_ci(est_time=1, suite="stage-a-unit-test-mlx") diff --git a/test/registered/unit/utils/test_field_validators.py b/test/registered/unit/utils/test_field_validators.py index 619586c632..d0a052cda4 100644 --- a/test/registered/unit/utils/test_field_validators.py +++ b/test/registered/unit/utils/test_field_validators.py @@ -7,7 +7,7 @@ from sglang.srt.utils.field_validators import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=11, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") class TestValidateListI64_1d(CustomTestCase): diff --git a/test/registered/unit/utils/test_gauge_histogram.py b/test/registered/unit/utils/test_gauge_histogram.py index 438b34973d..4e824f8b4b 100644 --- a/test/registered/unit/utils/test_gauge_histogram.py +++ b/test/registered/unit/utils/test_gauge_histogram.py @@ -1,6 +1,6 @@ from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=4, suite="stage-b-test-cpu-intel") import unittest diff --git a/test/registered/unit/utils/test_hf_transformers.py b/test/registered/unit/utils/test_hf_transformers.py index 92cc43cc37..c1f2b80dd8 100644 --- a/test/registered/unit/utils/test_hf_transformers.py +++ b/test/registered/unit/utils/test_hf_transformers.py @@ -31,7 +31,7 @@ from sglang.srt.utils.hf_transformers.tokenizer import _fix_special_tokens_patte from sglang.srt.utils.hf_transformers_patches import normalize_rope_scaling_compat from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # --------------------------------------------------------------------------- diff --git a/test/registered/unit/utils/test_hf_transformers_fastokens.py b/test/registered/unit/utils/test_hf_transformers_fastokens.py index 07dedb0476..7fae103776 100644 --- a/test/registered/unit/utils/test_hf_transformers_fastokens.py +++ b/test/registered/unit/utils/test_hf_transformers_fastokens.py @@ -12,7 +12,7 @@ from sglang.test.test_utils import ( TOKENIZER_MODEL = DEFAULT_SMALL_MODEL_NAME_FOR_TEST_QWEN -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=6, suite="base-a-test-cpu") try: diff --git a/test/registered/unit/utils/test_hf_transformers_loading.py b/test/registered/unit/utils/test_hf_transformers_loading.py index d41e047c70..aefebd272e 100644 --- a/test/registered/unit/utils/test_hf_transformers_loading.py +++ b/test/registered/unit/utils/test_hf_transformers_loading.py @@ -30,7 +30,7 @@ from sglang.srt.utils.hf_transformers import ( from sglang.srt.utils.patch_tokenizer import unpatch_tokenizer from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestHFTransformersLoading(unittest.TestCase): diff --git a/test/registered/unit/utils/test_invariants.py b/test/registered/unit/utils/test_invariants.py index 623233d94a..10a2759bb1 100644 --- a/test/registered/unit/utils/test_invariants.py +++ b/test/registered/unit/utils/test_invariants.py @@ -22,7 +22,7 @@ from sglang.srt.utils.invariants import ( from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") # "test." namespace so the coverage meta-test isolates these from real invariants. diff --git a/test/registered/unit/utils/test_json_response.py b/test/registered/unit/utils/test_json_response.py index 70aac292ce..9de70cc9d4 100644 --- a/test/registered/unit/utils/test_json_response.py +++ b/test/registered/unit/utils/test_json_response.py @@ -10,7 +10,7 @@ from sglang.srt.utils.json_response import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/utils/test_profile_merger.py b/test/registered/unit/utils/test_profile_merger.py index b8fcf87d0d..4ec0ba2ece 100644 --- a/test/registered/unit/utils/test_profile_merger.py +++ b/test/registered/unit/utils/test_profile_merger.py @@ -18,7 +18,7 @@ from sglang.srt.utils.profile_merger import ProfileMerger from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class TestProfileMerger(CustomTestCase): diff --git a/test/registered/unit/utils/test_subprocess_watchdog.py b/test/registered/unit/utils/test_subprocess_watchdog.py index cabae0756c..560bfb4598 100644 --- a/test/registered/unit/utils/test_subprocess_watchdog.py +++ b/test/registered/unit/utils/test_subprocess_watchdog.py @@ -24,7 +24,7 @@ from sglang.srt.utils.watchdog import SubprocessWatchdog from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase -register_cpu_ci(est_time=14, suite="base-a-test-cpu") +register_cpu_ci(est_time=8, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") diff --git a/test/registered/unit/utils/test_weight_versions.py b/test/registered/unit/utils/test_weight_versions.py index 7c2508b4b7..9bde6f3399 100644 --- a/test/registered/unit/utils/test_weight_versions.py +++ b/test/registered/unit/utils/test_weight_versions.py @@ -21,7 +21,7 @@ from sglang.test.test_utils import ( published_topology, ) -register_cpu_ci(est_time=12, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") class _ReqStub: diff --git a/test/registered/utils/test_log_utils.py b/test/registered/utils/test_log_utils.py index fab48d150c..f5715cbf29 100644 --- a/test/registered/utils/test_log_utils.py +++ b/test/registered/utils/test_log_utils.py @@ -10,7 +10,7 @@ from pathlib import Path from sglang.srt.utils.log_utils import create_log_targets, log_json from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") register_cpu_ci(est_time=4, suite="stage-b-test-cpu-intel") _LOG_PREFIX_RE = re.compile(r"^\[\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}\] ") diff --git a/test/registered/utils/test_network_address.py b/test/registered/utils/test_network_address.py index d3ca555b0d..0c4205393c 100644 --- a/test/registered/utils/test_network_address.py +++ b/test/registered/utils/test_network_address.py @@ -6,7 +6,7 @@ from sglang.srt.server_args import PortArgs, ServerArgs from sglang.srt.utils.network import NetworkAddress, is_zmq_endpoint_ipv6 from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") # Mock get_device() so ServerArgs tests run on CPU-only CI runners diff --git a/test/registered/utils/test_numa_utils.py b/test/registered/utils/test_numa_utils.py index ecc29d4658..337fa00be7 100644 --- a/test/registered/utils/test_numa_utils.py +++ b/test/registered/utils/test_numa_utils.py @@ -22,9 +22,9 @@ from sglang.srt.utils.numa_utils import ( ) from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") -register_cuda_ci(est_time=7, stage="base-c", runner_config="4-gpu-gb300") -register_cuda_ci(est_time=6, stage="base-c", runner_config="4-gpu-b200") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") +register_cuda_ci(est_time=3, stage="base-c", runner_config="4-gpu-gb300") +register_cuda_ci(est_time=3, stage="base-c", runner_config="4-gpu-b200") class TestIsNumaAvailable(unittest.TestCase): diff --git a/test/registered/utils/test_phase_checker.py b/test/registered/utils/test_phase_checker.py index d7bf6a12d5..d12d2a623f 100644 --- a/test/registered/utils/test_phase_checker.py +++ b/test/registered/utils/test_phase_checker.py @@ -19,7 +19,7 @@ from sglang.test.ci.ci_register import ( ) from sglang.test.test_utils import CustomTestCase -register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-small") +register_cuda_ci(est_time=16, stage="base-b", runner_config="1-gpu-small") register_amd_ci(est_time=120, stage="stage-b", runner_config="1-gpu-small-amd") # Nightly, not a blocking lane: one case spawns a subprocess that trips a device-side # assert, so a wedge costs the whole subprocess timeout below. diff --git a/test/registered/utils/test_socket_utils.py b/test/registered/utils/test_socket_utils.py index 6119b9a6d0..7dea1ea5c1 100644 --- a/test/registered/utils/test_socket_utils.py +++ b/test/registered/utils/test_socket_utils.py @@ -15,7 +15,7 @@ from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase from sglang.utils import normalize_base_url, release_port, reserve_port -register_cpu_ci(est_time=10, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=8, suite="stage-b-test-cpu-intel") diff --git a/test/registered/utils/test_stale_shm_cleanup.py b/test/registered/utils/test_stale_shm_cleanup.py index ef8b602344..da9925e8ab 100644 --- a/test/registered/utils/test_stale_shm_cleanup.py +++ b/test/registered/utils/test_stale_shm_cleanup.py @@ -12,7 +12,7 @@ from sglang.srt.utils.stale_shm_cleanup import ( ) from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=6, suite="base-a-test-cpu") +register_cpu_ci(est_time=1, suite="base-a-test-cpu") def _spawn_dead_pid() -> int: diff --git a/test/registered/vlm/test_video_utils.py b/test/registered/vlm/test_video_utils.py index 1ad3204f94..9c2517acea 100644 --- a/test/registered/vlm/test_video_utils.py +++ b/test/registered/vlm/test_video_utils.py @@ -5,7 +5,7 @@ import pytest from sglang.srt.utils import sample_video_frames from sglang.test.ci.ci_register import register_cpu_ci -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=2, suite="base-a-test-cpu") register_cpu_ci(est_time=5, suite="stage-b-test-cpu-intel") diff --git a/test/registered/vlm/test_vision_openai_server_extra.py b/test/registered/vlm/test_vision_openai_server_extra.py index 2e50ddbdc4..fc11f1503a 100644 --- a/test/registered/vlm/test_vision_openai_server_extra.py +++ b/test/registered/vlm/test_vision_openai_server_extra.py @@ -6,7 +6,7 @@ import unittest from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.vlm_utils import OmniOpenAITestMixin -register_cuda_ci(est_time=159, stage="extra-a", runner_config="1-gpu-large") +register_cuda_ci(est_time=382, stage="extra-a", runner_config="1-gpu-large") class TestQwen3OmniServer(OmniOpenAITestMixin): diff --git a/test/registered/xpu/test_numa_utils_xpu.py b/test/registered/xpu/test_numa_utils_xpu.py index 9d1378dd18..dd0755ebb7 100644 --- a/test/registered/xpu/test_numa_utils_xpu.py +++ b/test/registered/xpu/test_numa_utils_xpu.py @@ -23,7 +23,7 @@ from sglang.srt.utils.numa_utils import ( from sglang.test.ci.ci_register import register_cpu_ci, register_xpu_ci # CPU too: only TestXpuOrderMatchesRuntime needs hardware. -register_cpu_ci(est_time=7, suite="base-a-test-cpu") +register_cpu_ci(est_time=0, suite="base-a-test-cpu") register_xpu_ci(est_time=10, suite="stage-a-test-1-gpu-xpu")