Pleaplusone dc937175d4

[ROCm][Perf] New design on ROCm AITER MHA backend Implementation (#25763) ...

Signed-off-by: ganyi <ygan@amd.com>

2025-11-04 18:05:33 +00:00

..

attention

[ROCm][Perf] New design on ROCm AITER MHA backend Implementation (#25763)

2025-11-04 18:05:33 +00:00

core

[AsyncScheduling] Don't schedule past request max_tokens (#27922)

2025-11-04 17:06:28 +00:00

engine

[Core] Enable StatLogger in LLMEngine (#28020)

2025-11-04 04:13:35 -08:00

executor

[PerfFix] Avoid separate thread for MP executor shm spin (#28012)

2025-11-04 08:33:55 -08:00

kv_offload

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

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

metrics

[Metrics] Enable sleep state metric outside of dev mode (#27867)

2025-11-03 20:35:36 -08:00

pool

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

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

sample

[BugFix] Fix mixed penalties batch with async scheduling (#27910)

2025-11-01 10:51:24 -07:00

spec_decode

[model] Add support for openPangu_Ultra_MoE (#27521)

2025-11-04 08:17:20 -08:00

structured_output

[Core] Async scheduling + structured outputs compatibility (#26866)

2025-11-01 00:35:04 +00:00

worker

[PerfFix] Avoid separate thread for MP executor shm spin (#28012)

2025-11-04 08:33:55 -08:00

__init__.py

…

cudagraph_dispatcher.py

[LoRA] LoRA cuda graph specialization (#25914)

2025-10-20 04:21:09 +00:00

kv_cache_interface.py

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

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

outputs.py

[AsyncScheduling] Make async overlap work with logprobs (#27615)

2025-10-28 22:35:54 +00:00

request.py

…

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