[Bugfix] Fix: add patch_rope_scaling after hf override (#20857)

Wang Siyuan committed over 1 year ago

Support for LlamaForSequenceClassification (#20807)

Minkyu Kim committed over 1 year ago

Renable google/gemma-3-1b-it accuracy test. (#20866)

QiliangCui committed over 1 year ago

[Core] Support multiple tasks per model (#20771)

Nicolò Lucchesi committed over 1 year ago

[Refactor] Change the way of import triton (#20774)

Wentao Ye committed over 1 year ago

[Bugfix] Fix a couple PPLX+CUTLASS MoE bugs (#20825)

ElizaWszola committed over 1 year ago

[Perf] Use Triton instead of Torch for DeepGEMM Per Token Group Quant (#20841)

Wentao Ye committed over 1 year ago

[Sched] Enhance the logic to remove stopped requests from queues (#20739)

Woosuk Kwon committed over 1 year ago

[Bugfix] Restrict Machete to only run on Hopper (#20830)

Michael Goin committed over 1 year ago

[docs] convert supported configs to table (#20858)

Reid committed over 1 year ago

[Bugfix] Fix Tensor Parallelism Padding Consistency in Granite Models (#20843)

Alex Brooks committed over 1 year ago

[Model] New model support for microsoft/Phi-4-mini-flash-reasoning (#20702)

Congcong Chen committed over 1 year ago

Revert "Use NVCC --compress-mode to reduce binary size by 30% #20694" (#20853)

Michael Goin committed over 1 year ago

Enable ModelOpt Llama4 fp8 checkpoint deployment (#20419)

Zhiyu committed over 1 year ago

Remove extra tensor on CPU (#20693)

Maximilien de Bayser committed over 1 year ago

[cold start time] add envs.VLLM_COMPILE_DEPYF to guard decompile (#20790)

Boyuan Feng committed over 1 year ago

[Bugfix] Fix torch.compile x LoRA for PyTorch 2.8 (#20823)

Richard Zou committed over 1 year ago

[Docs] Update basic.md (#20846)

Lucia Fang committed over 1 year ago

[Bug] Fix DeepGemm for EP low latency case (#20833)

Wentao Ye committed over 1 year ago

[Misc] Respect `no_use_tqdm_on_load` flag while capturing CUDA graph (#20834)

lkchen committed over 1 year ago