[Deprecation] Remove dead use_prefill_decode_attention flag (#52557)

Brian Woods committed 2 months ago

[Model] Add Qwen3-Omni DSpark support (#52560)

Zhou248 committed 2 months ago

[Pooling] Fix Rust pooling endpoint for benchmark (#53352)

Taneem Ibrahim committed 2 months ago

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[eE][fF]" (#53381)

Wentao Ye committed 2 months ago

[Core][RL] Support sparse checkpoint updates through native weight loaders (#50723)

Shuolei Wang committed 2 months ago

[Perf] Support internal prefill checkpoints for Mamba prefix caching, 9%~25% TTFT improvement (#52789)

Wentao Ye committed 2 months ago

[MM] Remove renderer_applies_updates flag (#53385)

Cyrus Leung committed 2 months ago

[ROCm] Ship rocprofiler-sdk 1.3.2 in Dockerfile.rocm_base to fix torch.profiler traces (#53182)

Rohan Potdar committed 2 months ago

[Bugfix] Fix speculative decoding for short_conv (LFM2) models (#50272)

Jens committed 2 months ago

[Pooling] Report input throughput for batched requests (#53213)

Taneem Ibrahim committed 2 months ago

[CI][ARM64] Restore known-good CUDA 13 builder image (#53374)

Kevin H. Luu committed 2 months ago

[MM] Address comments on #53275 (#53364)

Cyrus Leung committed 2 months ago

[MM] Simplify _apply_hf_processor_main (#53275)

Cyrus Leung committed 2 months ago

[Refactor][Model Runner V2][Multimodal] Move the encoder-only path out of the shared runner (#53176)

Tianyu Guo committed 2 months ago

[Bugfix][Kimi K3] Enable deferred MoE finalization before weight loading (#53327)

Yongye Zhu committed 2 months ago

[Bugfix][R3] Unwrap UniformTypeKVCacheSpecs when selecting the routed-experts KV group (#53240)

ℍ𝕠𝕝𝕝𝕠𝕨 𝕄𝕒𝕟 committed 2 months ago

[Bugfix][ROCM] Fix the MXFP8 block scale exponent (#53110)

stefankoncarevic committed 2 months ago

[6/N][KV-Cache Layout Refactor] Standardize KV cache layout (#51718)

Lucas Wilkinson committed 2 months ago

[Bugfix][Spec Decode] Use group geometry for FlashAttention metadata (#53002)

Misha Goin committed 2 months ago

[Test] Add focused hybrid MTP prefix-cache regressions (#53189)

Misha Goin committed 2 months ago