Benjamin Chislett 304419576a

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479) ...

Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>

2025-11-13 01:56:40 +09:00

..

attention

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

2025-11-13 01:56:40 +09:00

core

[Core] Encoder separation for Encode-Prefill-Decode Disaggregation (#25233)

2025-11-11 18:58:33 -08:00

engine

[Perf] Use np.ndarray instead of list[list[int]] to reduce GC overhead (#28245)

2025-11-11 10:28:47 -08:00

executor

[Misc] Cleanup Executor interface (#28441)

2025-11-11 22:28:07 +00:00

kv_offload

[KV offload] Enable CPU KV offload on CUDA alike Platforms (#27770)

2025-10-30 22:10:29 +08:00

metrics

[Metrics] Refactor LoRA state tracking (#26801)

2025-11-10 16:34:36 +08:00

pool

[Chore] Separate out vllm.utils.platform_utils.py (#27374)

2025-10-23 19:08:06 +00:00

sample

[Performance] Cache loaded custom logitsprocs to avoid overheads (#28462)

2025-11-11 16:49:29 -08:00

spec_decode

[Attention] Refactor CUDA attention backend selection logic (#24794)

2025-11-11 07:40:44 -05:00

structured_output

[Bugfix] Fix llguidance backend, rollback when EOS was encountered (#25905)

2025-11-10 14:53:32 -05:00

worker

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

2025-11-13 01:56:40 +09:00

__init__.py

…

cudagraph_dispatcher.py

…

kv_cache_interface.py

[Hybrid] Pass kernel block size to builders (#27753)

2025-11-03 05:48:03 +00:00

outputs.py

[Core] Encoder separation for Encode-Prefill-Decode Disaggregation (#25233)

2025-11-11 18:58:33 -08:00

request.py

[Feature]: Add corrupted request metric to V1 metrics system. (#27306)

2025-11-05 13:45:29 -08:00

serial_utils.py

[Misc] Clean up more utils (#27567)

2025-10-27 15:30:38 +00:00

utils.py

[Misc] Clean up more utils (#27567)

2025-10-27 15:30:38 +00:00