[Bugfix][Frontend] Emit a content block for empty Anthropic completions (#46525)

Maxwill Lin committed 4 months ago

[Feat] Add runtime monitor for post-warmup CuTeDSL compilation (#46167)

Roberto L. Castro committed 4 months ago

[ROCm][Bugfix][Perf] enable shared expert fusion for Qwen3.5 (#44434)

Nico Holmberg committed 4 months ago

[Perf][DSv4/DSv3.2] Add cluster-cooperative topK kernel for low-latency scenarios (#43008)

Roberto L. Castro committed 4 months ago

[Perf] Skip detokenization in online beam search (#46422)

Guy Stone committed 4 months ago

[Optimization] Skip DP padding tokens in MoE (#46428)

Woosuk Kwon committed 4 months ago

[CI] Fix `test_auto_gptq` on ROCm CI (#46164)

fxmarty-amd committed 4 months ago

[ROCm][CI] Shard LM Eval Qwen3-5 Models (B200-MI355) in AMD CI (#46520)

Micah Williamson committed 4 months ago

[CI Test] Mark batch invariance test flaky (#46530)

Wentao Ye committed 4 months ago

[Bugfix][Frontend] Emit non-ASCII tool-call arguments without \uXXXX escapes (#46308)

Maxwill Lin committed 4 months ago

[Bugfix][ToolParser] Handle braces in required tool streaming strings (#45389)

Ting SUN committed 4 months ago

[Kernel] Add FlashInferCutedslMxfp8LinearKernel (cute-dsl mm_mxfp8) (#46393)

Yongye Zhu committed 4 months ago

[Bugfix] Allow flashinfer_cutlass as a clamped NVFP4 MoE backend (#46492)

Gabriel Wu committed 4 months ago

[NVFP4 MoE/Deepseek V4] Marlin: wire SwiGLU clamp + allow it for clamped models on non-Blackwell (#45836)

Mike G committed 4 months ago

[Docker] Remove redundant flashinfer download-cubin step (#46517)

Michael Goin committed 4 months ago

[Model Runer V2][DFlash] Fix lm head sharing for dflash (#46435)

Giancarlo Delfin committed 4 months ago

Deprecate old FP8 online MoE quantization class (#44514)

Yan Ma committed 4 months ago

[Bugfix][ROCm] Fix cumem sleep and teardown (#46203)

peizhang56 committed 4 months ago

[Kernel] Extend Marlin thread-tile padding to MoE (WNA16 + FP8/MXFP8) (#45703)

Michael Goin committed 4 months ago

[Bugfix] Parse MiniMax M3 streaming reasoning by text markers (#45718)

Rui Yin committed 4 months ago