Fix range_ratio Bug in RandomDataset (#16126)

WWW committed over 1 year ago

[TPU][V1] Disable per-request seed/Generator (#16172)

Nicolò Lucchesi committed over 1 year ago

[Bugfix] Fix output token length check logic (#16419)

look committed over 1 year ago

[VLM] Avoid unnecessary dummy multimodal data during processing (#16416)

Cyrus Leung committed over 1 year ago

[V1] Zero-copy tensor/ndarray serialization/transmission (#13790)

Nick Hill committed over 1 year ago

[Model] Reduce redundant computations in mamba2 blocks for Bamba-9B (#15423)

Chih-Chieh Yang committed over 1 year ago

[Bugfix] Fix bug when dataset is json (#15899)

Chenyaaang committed over 1 year ago

[V1][Spec Decode] Eagle Model loading (#16035)

Lily Liu committed over 1 year ago

[V1] Set structured output backend to `auto` by default (#15724)

Russell Bryant committed over 1 year ago

Improve configs - `ParallelConfig` (#16332)

Harry Mellor committed over 1 year ago

[TPU][V1] Use `language_model` interface for getting text backbone in MM (#16410)

Nicolò Lucchesi committed over 1 year ago

[VLM] Remove `BaseProcessingInfo.get_mm_max_tokens_per_item` (#16408)

Cyrus Leung committed over 1 year ago

Fix the torch version parsing logic (#15857)

Lu Fang committed over 1 year ago

[CI] Add auto update workflow for Dockerfile graph (#11879)

wineandchord committed over 1 year ago

[doc] update the wrong link (#16401)

Reid committed over 1 year ago

[Model] Remove image mm limit for LLaMa4 (#16365)

Ye (Charlotte) Qi committed over 1 year ago

Update Numba to 0.61.2 (#16376)

cyyever committed over 1 year ago

[doc] add download model tips (#16389)

Reid committed over 1 year ago

[Kernel] Use moe_wna16 kernel for compressed tensors wna16 moe models (#16038)

Michael Goin committed over 1 year ago

[Bugfix] Fix validation error for text-only Mllama 3.2 (#16377)

Cyrus Leung committed over 1 year ago