sglang

attention

Avoid materializing GDN QKV tensors during target verification (#33778 )

2026-09-21 17:04:18 -07:00

moe

Deprecate the parallel getters the context answers, and ratchet them shut (#40342 )

2026-09-21 12:25:32 -07:00

quantization

[ModelOpt][PP] Keep BF16 shared experts out of the NVFP4 fusion so TP1 pipeline stages can load (#40628 )

2026-09-21 21:45:58 -07:00

test_attn_residual.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_bailing_mrope_shape.py

Add Ling-3.0-flash-VL model support (#38526 )

2026-09-17 00:19:26 +08:00

test_conv_layer.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_dsv4_kv_splits_heuristic.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_dsv4_nonpaged_indexer.py

[DSV4] Chunk the indexer MQA logits by query rows under a free-memory budget (#39095 )

2026-09-19 11:53:20 -07:00

test_dsv41_candidate_blocks.py

[DeepSeek-V4.1] Bound dense prefill indexer memory (#40217 )

2026-09-20 14:43:19 -07:00

test_flashattention_paged_mha.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_flashinfer_comm_fusion.py

Use fp32 in TRTLLM all reduce buffers (#36143 )

2026-09-07 16:57:07 +08:00

test_fp8_bpreshuffle_dense_linear_mi35x.py

[AMD] DeepSeek-V4 MI355X: eliminate bpreshuffle fp8-scale relayout copy in dense w8a8 linear (#33165 )

2026-08-19 03:02:40 -07:00

test_fp8_bpreshuffle_producer_mi35x.py

[AMD] DeepSeek-V4 MI355X: eliminate bpreshuffle fp8-scale copies at producer sites (MoE down, MLA o_proj bmm) (#33166 )

2026-08-20 21:46:23 -07:00

test_fp8_bpreshuffle_scale.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_gdn_mis_metadata.py

Fix Qwen3.5 GDN multi-item scoring (#33922 )

2026-09-09 22:30:39 -07:00

test_kda_decode_mtp_slot_stride.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_kpool_planner_cpu_mirror.py

[GLM-5.3-Flash] Reduce KPool planning synchronization and overlap indexer preparation (#39695 )

2026-09-19 23:51:17 -07:00

test_kpool_stream_scheduling.py

[GLM-5.3-Flash] Reduce KPool planning synchronization and overlap indexer preparation (#39695 )

2026-09-19 23:51:17 -07:00

test_layer_communicator_fusion_gate.py

[Test] Fix fusion-group mocks after runtime context migration (#40290 )

2026-09-19 14:45:16 +08:00

test_layer_scatter_modes_cp_dense_mlp.py

[CP V1 Deprecation 4/5] Canonicalize prefill CP API names (#36229 )

2026-09-08 16:03:36 -07:00

test_layernorm_sp.py

[Config] Round 6.4: the runtime reads the bags, not the record (#38049 )

2026-09-06 21:41:46 -07:00

test_logprob_chunk_stitching.py

[Logprob] Serve input-logprob temporaries from CUDA-graph-pool dead space (#40038 )

2026-09-19 12:03:52 -07:00

test_logprob_fast_input.py

[Logprob] Serve input-logprob temporaries from CUDA-graph-pool dead space (#40038 )

2026-09-19 12:03:52 -07:00

test_mamba2_track_ssm_indices.py

Fuse GLM-5.3-Flash KDA projections and prefill metadata (#39688 )

2026-09-19 23:46:33 -07:00

test_mamba_prefill_track_metadata.py

Retire the per-runner parallel record (#40343 )

2026-09-21 12:26:40 -07:00

test_mamba_state_scatter_triton.py

[XPU][Fix] Pack device-pointer tables as uint64 to avoid 64-bit address overflow (#35051 )

2026-09-10 10:12:18 +08:00

test_minicpm_attention_adapter.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_minicpm_sparse_cache.py

[Config] Round 6.5: a namespace declares what it derives, next to what it derives it from (#38113 )

2026-09-06 21:44:24 -07:00

test_minicpm_sparse_metadata.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_moriep_mxfp8_dispatch.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_mova.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_pooler_score_and_pool.py

model: support EmbeddingGemma (#32375 )

2026-07-27 10:40:47 +08:00

test_radix_attention.py

Carry deferred attention operands and reuse multimodal shared memory (#39870 )

2026-09-17 01:17:15 -07:00

test_radix_linear_attention.py

chore: update CI test est_time values (#38238 )

2026-09-06 17:49:41 -07:00

test_yarn_cache_extension.py

[Fix] Preserve YaRN scaling when extending rotary caches (#38786 )

2026-09-21 14:14:05 +08:00