Jialin Ouyang ccd98b59c1

[Perf] Introduce FlattenLogprobs to store logprobs results to reduce GC overhead (#28171) ...

Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>

2025-11-07 00:27:12 -08:00

..

attention

[FlashInfer] Avoid FlashInfer block_size 16 + head_size 256 on blackwell (#27994)

2025-11-05 09:25:32 -08:00

core

[Core][Hybrid allocator + connector 2/n] Unify remove_skipped_blocks by get_last_useful_token (#25431)

2025-11-06 00:12:00 +00:00

engine

[Perf] Introduce FlattenLogprobs to store logprobs results to reduce GC overhead (#28171)

2025-11-07 00:27:12 -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

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

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

pool

…

sample

[Bugfix] Validate custom logits processor xargs for online serving (#27560)

2025-11-05 16:53:33 +00:00

spec_decode

[BugFix] Support EP/DP + EPLB with MTP (#25311)

2025-11-05 15:22:17 +00:00

structured_output

[Core] add support for reasoning parser plugins (#28075)

2025-11-06 01:15:06 +08:00

worker

[HARDWARE][CPU] Add Option for Disabling Binding to Specific CPU Cores (#27953)

2025-11-06 23:47:11 +08:00

__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

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

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

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