[TPU] Add TPU specific var VLLM_TPU_MOST_MODEL_LEN (#19919)

Chenyaaang committed over 1 year ago

[P/D] Avoid stranding blocks in P when aborted in D's waiting queue (#19223)

Nick Hill committed over 1 year ago

[TPU][Bugfix] fix kv cache padding (#20048)

Chengji Yao committed over 1 year ago

[Bugfix] default set cuda_graph_sizes to max_num_seqs for v1 engine (#20062)

zhrrr committed over 1 year ago

Fix the path to the testing script. (#20082)

QiliangCui committed over 1 year ago

[Misc] Add parallel state `node_count` function (#20045)

Nick Hill committed over 1 year ago

[PD] let toy proxy handle /chat/completions (#19730)

lkchen committed over 1 year ago

static_scaled_fp8_quant should not run when scale.numel is not 1 (#20076)

Eldar Kurtić committed over 1 year ago

[Fix] Support cls pooling in ModernBertPooler (#20067)

lsz05 committed over 1 year ago

[Quantization] Add compressed-tensors emulations support for NVFP4 (#19879)

Dipika Sikka committed over 1 year ago

[Frontend] Add `/v1/audio/translations` OpenAI API endpoint (#19615)

Nicolò Lucchesi committed over 1 year ago

[V1][Speculative Decoding] Fix DeepSeek MTP (#20022)

cjackal committed over 1 year ago

[Doc] Guide for Incremental Compilation Workflow (#19109)

Michael Goin committed over 1 year ago

[doc] add reference link for Intel XPU (#20064)

Reid committed over 1 year ago

[BugFix] Fix full-cuda-graph illegal memory access in FA3 (#20057)

Lucas Wilkinson committed over 1 year ago

[Chore] debloat some initial logs (#19438)

Aaron Pham committed over 1 year ago

[Kernels][Bugfix] Use torch op for all kernels in FusedMoE forward. Add additional testing for cudagraphs. (#19717)

bnellnm committed over 1 year ago

[Feat][CLI] enforce-include-usage (#19695)

Max Wittig committed over 1 year ago

[Refactor] Remove duplicate `ceil_div` (#20023)

Wentao Ye committed over 1 year ago

[Frontend] speed up import time of vllm.config (#18036)

David Xia committed over 1 year ago