Files

T

liuzhenweiandKunshang Ji 6654bcbfca [MRV2][XPU] use xpu sample kernel in mrv2 sampler (#57277 )

Signed-off-by: zhenwei-intel <zhenwei.liu@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

2026-09-21 13:24:41 +08:00

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

[Feat][Model] Support encoder-side SWA-bounded replay for DeepSeek-V4.1-Flash (#56227 )

2026-09-18 15:01:24 -07:00

pool

[Pooling] Support prompt embeddings in MRV2 decoder pooling (#56639 )

2026-09-13 05:49:45 +00:00

sample

[MRV2][XPU] use xpu sample kernel in mrv2 sampler (#57277 )

2026-09-21 13:24:41 +08:00

spec_decode

[Bugfix][Spec Decode] Stop dummy draft decode steps from writing KV through stale block-table rows (#56734 )

2026-09-21 01:06:25 +00:00

__init__.py

…

async_utils.py

[Perf][Model Runner V2] Compact sampling masks on GPU instead of unpacking the full-vocab bitmask on CPU (#54901 )

2026-09-03 19:03:36 -07:00

attn_utils.py

[Bugfix][MRV2] Match fast-prefill padding to active LoRA batches (#56456 )

2026-09-18 22:13:23 +00:00

block_table.py

[Bugfix][Spec Decode] Stop dummy draft decode steps from writing KV through stale block-table rows (#56734 )

2026-09-21 01:06:25 +00:00

buffer_utils.py

[MRV2] Buffer util simplifications (#56888 )

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

cp_utils.py

[Bugfix][MRV2] Initialize DCP metadata after batch partitioning (#55212 )

2026-09-10 06:50:09 +00:00

cudagraph_utils.py

[Model] Pass intermediate_tensors to the model when capturing CUDA gr… (#57745 )

2026-09-19 23:23:29 -07:00

dp_utils.py

Add pydocstyle to the ruff rules (#52136 )

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

ec_connector.py

[EC Connector] Add Metrics Collection (#54960 )

2026-09-17 10:36:23 +08:00

eplb_utils.py

[Elastic EP] Reduce eager-mode reconfiguration downtime (#51885 )

2026-08-19 02:13:06 +00:00

input_batch.py

[Bugfix][Model Runner V2] Route dummy tokens to MoE experts during profiling (#57270 )

2026-09-17 10:11:14 -04:00

kv_connector.py

[3/N] HiSparse: host-resident sparse-MLA decode hot-buffering (#53781 )

2026-09-12 00:36:47 -04:00

lora_utils.py

Add pydocstyle to the ruff rules (#52136 )

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

model_runner.py

[Perf] Give a prefill-only batch the model state's number of logit rows (#57416 )

2026-09-20 04:58:30 +00:00

pcp_manager.py

[MRV2] Buffer util simplifications (#56888 )

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

pp_utils.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212 )

2026-09-16 10:57:24 -07:00

README.md

…

shutdown.py

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

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

states.py

[Model Runner V2][Spec Decode] Add multi-layer MTP speculator (#48892 )

2026-07-30 15:54:15 -07:00

structured_outputs.py

[MRV2] Buffer util simplifications (#56888 )

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

ubatch_utils.py

[2/2][Model Runner V2] FULL CUDA graph capture for microbatched steps (DBO) (#51700 )

2026-09-13 22:42:33 -07:00

warmup.py

[3/N] HiSparse: host-resident sparse-MLA decode hot-buffering (#53781 )

2026-09-12 00:36:47 -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.