[Benchmark] Add retry support to fix workload bias in multi-turn benchmark (#28493)

ai-jz committed about 1 year ago

[BugFix] Add fallback path in `apply_rotary_pos_emb_flashattn` for non-cuda platforms (#28447)

Fanli Lin committed about 1 year ago

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

Chenguang Zheng committed about 1 year ago

[Model][Qwen3VL] Simplify `get_mrope_input_positions` using numpy (#28302)

Lukas Geiger committed about 1 year ago

`VLLM_USE_TRITON_FLASH_ATTN` V0 variable deprecation (#27611)

Andreas Karatzas committed about 1 year ago

[CPU] Refactor CPU attention backend (#27954)

Li, Jiang committed about 1 year ago

[[V0 deprecation]]Remove VLLM_USE_V1 env (#28204)

wangxiyuan committed about 1 year ago

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

Isotr0py committed about 1 year ago

[Frontend] Change CompilationMode to a proper Enum (#28165)

Yanan Cao committed about 1 year ago

[BugFix] Graceful handling of torch symm mem errors. (#27671)

Ilya Markov committed about 1 year ago

Use FLASHINFER MLA backend when testing fp8_kv_scale_compile (#28491)

Adrian Abeyta committed about 1 year ago

[Feature] Allow configuring FlashInfer workspace size (#28269)

Max Hu committed about 1 year ago

[Bugfix] Disable shared expert overlap if Marlin MoE is used (#28410)

Michael Goin committed about 1 year ago

Add Zurich vLLM Meetup (#28488)

Michael Goin committed about 1 year ago

[Misc] Cleanup Executor interface (#28441)

wangxiyuan committed about 1 year ago

[TPU] Rename path to tpu platform (#28452)

Kyuyeun Kim committed about 1 year ago

[LoRA][1/N]Remove LoRA extra vocab (#28382)

Jee Jee Li committed about 1 year ago

[Misc] fix typo in DCP comment (#28389)

Jie Luo committed about 1 year ago

[V0 deprecation] Clean up num_prefill_tokens logic for V0 (#28203)

Canlin Guo committed about 1 year ago

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

Jialin Ouyang committed about 1 year ago