Woosuk Kwon 6049424b7e [MRV2] Always build attn metadata at capture time (#49364) ...

Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>

2026-07-21 19:27:16 -07:00

..

cpu

Revert "[Platform] Replace torch.cuda.Event with torch.Event (#47140)" (#47668)

2026-07-06 14:18:33 +01:00

gpu

[MRV2] Always build attn metadata at capture time (#49364)

2026-07-21 19:27:16 -07:00

__init__.py

…

block_table.py

[Core] Add MRV2 virtual-batch PCP for MLA (#46570)

2026-07-19 02:53:15 +00:00

cp_utils.py

[Core] Add MRV2 virtual-batch PCP for MLA (#46570)

2026-07-19 02:53:15 +00:00

cpu_model_runner.py

[CPU] Remove speculative decoding stream overrides from CPUModelRunner (#47162)

2026-07-01 06:12:49 +00:00

cpu_worker.py

[CPU] Create Proper Numa topology for s390x (#40714)

2026-07-13 12:58:43 +08:00

dp_utils.py

[Perf][1/n] Eliminate various GPU<->CPU syncs (#41429)

2026-05-11 20:36:03 +00:00

ec_connector_model_runner_mixin.py

[Refactor] Remove dead code in kv connector and model runner (#38383)

2026-03-31 17:05:23 -04:00

encoder_cudagraph_defs.py

[MM][Perf][CG] Support dual-path ViT full CUDA graph for DeepSeek-OCR (#43586)

2026-06-16 04:35:20 -07:00

encoder_cudagraph.py

[MM][Perf][CG] Support dual-path ViT full CUDA graph for DeepSeek-OCR (#43586)

2026-06-16 04:35:20 -07:00

gpu_input_batch.py

[Bugfix] Fix thinking_token_budget not enforced after natural </think> re-entry (#45984)

2026-07-10 22:47:51 +00:00

gpu_model_runner.py

[3/N][KV-Cache Layout Refactor] Standardize Mamba cache; drop get_transfer_cache_regions (#44456)

2026-07-21 09:16:15 +00:00

gpu_ubatch_wrapper.py

[Bugfix] Fix UBatchWrapper CUDA graph key to sum all ubatches, not just first two (#43161)

2026-07-07 12:42:31 +00:00

gpu_worker.py

[Misc] Move PyNvVideoCodec stuff out of gpu worker (#49322)

2026-07-22 10:11:56 +08:00

kv_connector_model_runner_mixin.py

[Spec Decode] Support mixed KV page sizes for DFlash (#45181)

2026-06-21 22:45:14 +08:00

lora_model_runner_mixin.py

[Bugfix] LoRA for DeepSeek V3.2 (#35077)

2026-04-22 19:33:50 +08:00

mamba_utils.py

[Perf][Hybrid] Vectorize _copy_mamba_state_block to uint64 for temporal (#48110)

2026-07-18 04:43:09 +00:00

startup_plan.py

[Core] Persist and reuse the memory-profiling result across boots (opt-in) (#47388)

2026-07-08 00:53:02 +00:00

tpu_input_batch.py

DCP supports hybrid attention (#40996)

2026-07-09 21:34:45 -07:00

ubatch_utils.py

[BugFix] Preserve max_seq_len in ubatch metadata during CUDA graph capture (#40961)

2026-05-05 04:29:34 +00:00

ubatching.py

[Misc] Replace Optional[X] with X | None syntax (#33332)

2026-01-30 01:56:59 -08:00

utils.py

[3/N][KV-Cache Layout Refactor] Standardize Mamba cache; drop get_transfer_cache_regions (#44456)

2026-07-21 09:16:15 +00:00

worker_base.py

Stop setting CUDA_VISIBLE_DEVICES internally in vLLM, add device_ids arg (#45026)

2026-06-20 13:38:10 -07:00

workspace.py

[Bugfix] Fix workspace resize leaking reserved GPU memory (#39226)

2026-04-23 20:50:05 +00:00

xpu_model_runner.py

[XPU] Fix Event init failure w/ blocking (#47868)

2026-07-07 22:54:03 +08:00

xpu_worker.py

[Doc] Fix grammatically incorrect error message in gpu_worker and xpu_worker (#36715)

2026-07-07 17:03:06 +00:00