YAMY 3c9cea8f10 [EAGLE] Prune draft-extend logits to selected rows (#35546)

2026-09-02 15:10:08 -07:00

..

model_runner_components

[Memory] Size the CUDA graph pool from warmup measurements and fix graph-pool borrowing (#36911)

2026-09-01 09:32:38 -07:00

runner

config: the resolution callbacks into the record go to zero (#36972)

2026-08-29 04:18:05 -07:00

runner_backend

[Memory] Size the CUDA graph pool from warmup measurements and fix graph-pool borrowing (#36911)

2026-09-01 09:32:38 -07:00

runner_utils

[Memory] Size the CUDA graph pool from warmup measurements and fix graph-pool borrowing (#36911)

2026-09-01 09:32:38 -07:00

test_chunked_prefix_cache_gate.py

config: the draft runner carries its own attention backend

2026-08-05 19:32:24 -07:00

test_cuda_graph_buffer_registry.py

[EAGLE] Prune draft-extend logits to selected rows (#35546)

2026-09-02 15:10:08 -07:00

test_draft_runner_skips_lora.py

[Spec][LoRA] Support multi-adapter LoRA with EAGLE/NEXTN/DFLASH/DSPARK speculative decoding (#34337)

2026-08-21 14:21:53 -07:00

test_forward_metadata_plan_record.py

Replace skip_attn_backend_init with a batch-carried attention plan marker (+ staleness re-plan) (#27193)

2026-06-04 17:13:18 -07:00

test_hisparse_pool_configurator.py

[DSA] Skip indexer KV cache for skip-topk layers (#30531)

2026-08-17 02:02:23 -07:00

test_mlp_sync_pad_unpad.py

[Feature] Coordinate FullCG prefill across DP-attention ranks (#35640)

2026-08-26 02:16:02 -07:00

test_model_runner_decode_rows.py

[Bug][Spec] fix startup crash and reduce CUDA graph memory usage for speculative adaptive (#35275)

2026-08-27 00:46:13 -07:00

test_pool_configurator.py

config: three cache and pool readers take the bags (#36791)

2026-08-28 10:21:32 -07:00

test_prefill_cuda_graph_runner_helpers.py

[PP] Support prefill CUDA graph proxy tensors (#36248)

2026-08-30 23:11:45 -07:00

test_prefill_cuda_graph_runner.py

[PP] Support prefill CUDA graph proxy tensors (#36248)

2026-08-30 23:11:45 -07:00

test_unified_out_cache_loc_rebind.py

[Feature] Unified memory: support decode context parallelism for Kimi-Linear (#36890)

2026-09-01 12:44:26 -07:00