[CI][Test] Deflake the rms_norm scaling-property assertions (#54271)

Oxana Korzh committed about 2 months ago

[Bugfix] Set breakable graph env before Ray actor import (#53293)

Ace Eldeib committed about 2 months ago

[ROCm][PERF] Enable AITER PA gluon decode for MiniMax-M3 MTP and dense layers (#52849)

ukannika committed about 2 months ago

[CI/Build] Add advisory PR title format check (#54263)

Bugen Zhao committed about 2 months ago

[Model] Support speculative decoding method for PLaMo3 (#54239)

Shinichi Hemmi committed about 2 months ago

[ROCm][CI] Fix test_ray_v2_executor (#54249)

Charlie Fu committed about 2 months ago

[Bugfix] Keep the Moondream3 MoE all-reduce out of the fused-path try (#54152)

Oxana Korzh committed about 2 months ago

[Core][KV Connector] Start async KV loads after the forward launch when no sync loads are scheduled (#53333)

Summer Yang committed about 2 months ago

[Bugfix][MRV2] Release layer-bound KV cache memory in shutdown() (#54246)

Nick Hill committed about 2 months ago

[Bugfix][ROCm] Pre-allocate `wvSplitKrc` static workspaces before KV init (#54247)

Nick Hill committed about 2 months ago

[Mypy] Fix mypy typing for model interfaces and H/I models (#54079)

Taneem Ibrahim committed about 2 months ago

[Rust Frontend] Optimize SSE streaming hot path (#51321)

Reid committed about 2 months ago

[Bugfix] Fix int32 token offset overflow in fused SiLU block quant (#53409)

canlahlah committed about 2 months ago

[BugFix] Bind RayExecutorV2 TCPStore before publishing its port (#50969)

aoshen02 committed about 2 months ago

[ROCm] remove VLLM_ROCM_USE_AITER_FP4_ASM_GEMM environment variable; make w4a4 use the preshuffle triton+asm by default (#53141)

afriedri committed about 2 months ago

update quark docs to include online quantization (#52736)

hangy-amd committed about 2 months ago

[Kimi-K3][Kernel] Optimize the low-M fused latent MoE tail (#54168)

Yongye Zhu committed about 2 months ago

[Perf] Split xdrope_positions H2D copy into per-row transfers (#53412)

Keyi Li committed about 2 months ago

[MM][CG] Support ViT full CUDA graph for Idefics3 and SmolVLM (#47625)

Chipmunk committed about 2 months ago

[ROCm][Quantization][MOE] Enable fused shared experts for block-quantized FP8 (#53097)

xuebwang-amd committed about 2 months ago