[Bugfix] Add Kimi K3 MoE support to benchmark_moe.py (#50082)

vanshbhatia-amd committed 2 months ago

[Bugfix][Quantization] Support CT block FP8 with Marlin (#52966)

Misha Goin committed 2 months ago

[CI] Harden RemoteVLLMServer GPU cleanup checks (#52282)

Andreas Karatzas committed 2 months ago

[ROCm][CI] Enable modular OAI Triton MoE tests (#46434)

Andreas Karatzas committed 2 months ago

Revert "[Kernel] Gemma-4 FA4 FP8 Kernel" (#52987)

Roger Wang committed 2 months ago

[CI/Build] Fix CPU platform pre-commit formatting (#52981)

Misha Goin committed 2 months ago

[LoRA] Avoid false target matches for unsupported module types (#52313)

linitra24 committed 2 months ago

[Bugfix][V1] Sync mamba_block_size via EngineCoreReadyResponse (#50809)

lxy committed 2 months ago

[ROCm] Give EngineCore cleanup grace after request abort (#52281)

Andreas Karatzas committed 2 months ago

[Model] Support bidirectional (encoder-only) attention for DeepSeek e… (#52948)

Egor committed 2 months ago

[CT] Support Humming for WNA16 MoE (#48918)

Yi Liu committed 2 months ago

[Bugfix][Quantization] Fix OCP MX MoE emulation silently skipping mxfp6 activation QDQ (#52704)

xuebwang-amd committed 2 months ago

[ROCm] [Bugfix] Fix Triton fused shared expert alignment (#51632)

akii96 committed 2 months ago

[Bugfix][CPU] Enable C++ causal_conv1d GDN path and float32 SSM cache on non-AMX AVX-512BF16 CPUs (#49688)

Dinesh Chitlangia committed 2 months ago

[ROCm] Add UE8M0 scale packing for Triton silu_mul_quant (#37835)

Andreas Karatzas committed 2 months ago

[ROCm][CI] Extended Fused MoE and FP8 MoE test support (#41100)

Andreas Karatzas committed 2 months ago

[Bugfix] compressed-tensors: restore int8 grouped WNA16 MoE support (#52002)

Yohann Prigent committed 2 months ago

[Build] Add InstantTensor to CUDA dependencies (#52801)

Misha Goin committed 2 months ago

[Core] Skip broadcasting mm tensor data to workers for prefix-cache-covered items (#52041)

Xiaoan (Sean) Liu committed 2 months ago

[Kernel] SM120: stop routing misaligned-M blockwise FP8 GEMMs to the small-M swapAB config (#52775)

Gabriel Wu committed 2 months ago