[CI] Support partial torch requirement contexts (#51832)

Taneem Ibrahim committed 2 months ago

[Bugfix] Generalize KV block zeroing to `AttentionSpec` (#51749)

Michael Goin committed 2 months ago

[Bugfix] Take the sliding window from the layer, not the KV cache group (#51756)

Nick Hill committed 2 months ago

Bump Transformers version to 5.15.0 (#51668)

Harry Mellor committed 2 months ago

[Bugfix][MRV2] Support encoder timing stats in model runner V2 (#50020)

Guan-Ming Chiu committed 2 months ago

[Kernel][ROCm][Perf] FlyDSL decode-attention kernel for 4-bit TurboQuant KV cache (#47896)

aditi-amd committed 2 months ago

[XPU][CI] Fix ExampleConnector KV cache device selection (#51806)

liuzhenwei committed 2 months ago

[Bugfix] Align Qwen GDN gates with speculative tokens (#51812)

Jiangyun Zhu committed 2 months ago

[Misc] Enable test_silu_mul_fp8_quant_deep_gemm on XPU (#49444)

pmanczak committed 2 months ago

[Config] Update default `_max_num_batched_tokens` from 8192 to 16384 (#51726)

Wentao Ye committed 3 months ago

[Bugfix][MoE] Support GELU tanh in FlashInfer B12x MoE (#51819)

Andrii Skliar committed 3 months ago

[Bugfix] Guard DeepSeek V4 MRV1 piecewise CUDA graphs (#51768)

Woosuk Kwon committed 3 months ago

Add MoE output contract for MoE tail fusion (#51407)

Jee Jee Li committed 3 months ago

[Bugfix][ROCm] Fix DeepSeek V4 DSpark probabilistic startup (#51145)

Tuukka Sarvi committed 3 months ago

Bound generation inputs before expensive work (#51447)

Clinton Thomas committed 3 months ago

connects vLLM Recipes with vLLM's native config-based deployment and benchmark (#51308)

Louie Tsai committed 3 months ago

[XPU] Fix UVA weight offloading (non-pinned-tensor views and static Triton launcher) (#51770)

Chaojun Zhang committed 3 months ago

[Bugfix][CPU] Make the Apple Silicon BF16 probe fall back instead of raising (#51627)

Kush Zingade committed 3 months ago

[2/N][Feat][Perf] Add new warmup infrastructure for JITs. Add predicate filtering for JIT warmup, and migrate Inkling FA4 (#49315)

Roberto L. Castro committed 3 months ago

[Model] Enable tower and connector LoRA for Keye (#51780)

liushujia122 committed 3 months ago