FP32 router GEMV optimization (#48335)

Jee Jee Li committed 4 months ago

[CI] Point CI at Transformers release rather than release branch (#48328)

Harry Mellor committed 4 months ago

[XPU]remove is_xxx from moe class and bump up kernels (#48079)

Qiming Zhang committed 4 months ago

Bump Transformers version to 5.13.0 (#47867)

Harry Mellor committed 4 months ago

[Revert] [Build] Update vllm ...builds FA3 with torch stable API (#48269)

Lucas Wilkinson committed 4 months ago

[Logs] DP Supervisor Log Improvement (#48278)

Robert Shaw committed 4 months ago

[Misc] Use meta tensor for KV cache stride calculation (#47316)

Lucas Wilkinson committed 4 months ago

[BugFix] Fix packed HND KV cache reshape for FlashAttention (#47314)

Lucas Wilkinson committed 4 months ago

[BugFix] weights processing peak memory reduction for nvfp4 MoE layers (#46276)

Jimmy Lee committed 4 months ago

[Quantization] Bound peak memory when repacking FP4 MoE weights for Marlin (#47851)

Joe Rowell committed 4 months ago

Add VLLM_FLASHINFER_AUTOTUNE_SKIP_OPS and skip CuTeDSL fp4_gemm autotuning by default (#48268)

Michael Goin committed 4 months ago

[Bugfix] Fix thinking_token_budget not enforced after natural </think> re-entry (#45984)

Ashwin Giridharan committed 4 months ago

[Bugfix][Test] Register Qwen/Qwen3.5-4B example model (#48276)

Nick Hill committed 4 months ago

handle topk_ids padding in align sum kernel (#47785)

gnovack committed 4 months ago

[Bugfix] Fix FlashMLA dense fp8 metadata crash (num_sm_parts clamp) (#48045)

Matthew Bonanni committed 4 months ago

[ROCm] Enable DeepSeek-V4 DSpark speculative decoding on AMD (MI350X / MI355X, gfx950) (#47419)

larryli2-amd committed 4 months ago

[Model] Add LongCat-Flash-Lite (n-gram embedding) (#47857)

Michael Goin committed 4 months ago

[Build/CI] Build arm64 PR and postmerge image builds for Blackwell SM10x and SM110 (#48041)

Tyler Michael Smith committed 4 months ago

fix: correct load_weights track logic and enable weight integrity for… (#41811)

HuYiPeng committed 4 months ago

[Bugfix] Fix turboquant FP8 cast failure for BF16 models on Ampere GPUs (#39988)

XuZhou committed 4 months ago