Files

T

eb2901b735 [Perf][DSv4.1] Run decoder replay layers in CUDA graphs and trim in PIECEWISE graphs (#59532 )

Signed-off-by: Juntian Liu <Juntianl777@gmail.com>
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

2026-10-05 22:11:45 -07:00

metrics

[Model Runner V2] Support num NaNs in logits (#30187 )

2025-12-09 10:00:49 -08:00

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

model_states

[Perf][DSv4.1] Run decoder replay layers in CUDA graphs and trim in PIECEWISE graphs (#59532 )

2026-10-05 22:11:45 -07:00

pool

[Bugfix][Pooling] Handle multimodal cache misses without crashing (#58975 )

2026-10-02 10:14:33 -04:00

sample

[Feature] Per-row candidate IDs for prefill token scoring (M2 of #56860 ) (#56984 )

2026-10-03 14:17:09 +00:00

spec_decode

[Bugfix][MRV2][Spec Decode] Honor dynamic K in autoregressive speculators, 1.2~1.3x kernel perf improvement (#57053 )

2026-10-05 14:54:11 -04:00

__init__.py

…

async_utils.py

[Feature][Spec Decode] Support sampling mask replay for MRV2 MTP (#59359 )

2026-10-01 13:51:39 +08:00

attn_utils.py

[CI] Bump Transformers version to 5.18.0 (#59621 )

2026-10-04 00:42:35 +00:00

block_table.py

[Elastic EP] Support Model Runner V2 (#53934 )

2026-09-29 00:20:40 +00:00

buffer_utils.py

[MRV2] Buffer util simplifications (#56888 )

2026-09-15 20:36:51 +00:00

cp_utils.py

[MRV2] Miscellaneous code cleanup (#57980 )

2026-09-23 01:51:52 -07:00

cudagraph_utils.py

[Bugfix][MRV2][Spec Decode] Honor dynamic K in autoregressive speculators, 1.2~1.3x kernel perf improvement (#57053 )

2026-10-05 14:54:11 -04:00

dp_utils.py

[Spec Decode] Enable Gemma4 DSpark adaptive verification with FlashInfer (#57263 )

2026-09-26 17:01:17 +08:00

ec_connector.py

[MRV2] Miscellaneous code cleanup (#57980 )

2026-09-23 01:51:52 -07:00

eplb_utils.py

[Elastic EP] Support Model Runner V2 (#53934 )

2026-09-29 00:20:40 +00:00

input_batch.py

[Perf][PP] Skip sampled-token broadcasts whose requests leave the engine (#58542 )

2026-09-30 19:18:55 -07:00

kv_connector.py

[Bugfix][KVConnector] Finalize saves on steps without a forward (#57775 )

2026-09-25 00:06:36 +00:00

lora_utils.py

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

model_runner.py

[Perf][DSv4.1] Run decoder replay layers in CUDA graphs and trim in PIECEWISE graphs (#59532 )

2026-10-05 22:11:45 -07:00

pcp_manager.py

[Perf][PCP] Shard decode requests across PCP ranks (#52162 )

2026-10-01 23:50:13 -04:00

pp_utils.py

[Perf][PP] Skip sampled-token broadcasts whose requests leave the engine (#58542 )

2026-09-30 19:18:55 -07:00

README.md

…

shutdown.py

[MRV2] Add shutdown() method (#41297 )

2026-05-02 21:06:30 -07:00

states.py

[MRV2] Miscellaneous code cleanup (#57980 )

2026-09-23 01:51:52 -07:00

structured_outputs.py

[Bugfix] Don't let a structured-output request sample from an unmasked row (#54442 )

2026-10-01 02:11:44 +00:00

ubatch_utils.py

[Perf][PP] Skip sampled-token broadcasts whose requests leave the engine (#58542 )

2026-09-30 19:18:55 -07:00

warmup.py

[Bugfix][MRV2][Spec Decode] Honor dynamic K in autoregressive speculators, 1.2~1.3x kernel perf improvement (#57053 )

2026-10-05 14:54:11 -04:00

README.md

[Experimental] Model Runner V2

This directory contains the new model runner which is under active development. Ping Woosuk Kwon for any changes.