[Bugfix] Cache added_vocab to avoid per-token overhead (#30351)

Mingliang Li committed 11 months ago

Fix typos in comments across multiple files (#30345)

Wilson Wu committed 11 months ago

[bugfix][quantization] fix quark qwen3 kv_cache quantization (#30308)

haoyangli-amd committed 11 months ago

[CI] Reduce Flakiness For test_spec_decode.py::test_suffix_decoding_acceptance (#30367)

Micah Williamson committed 11 months ago

[CI/Build][AMD] Skip quantization kernels tests that require CUTLASS or e4m3fn when not supported by platform (#30020)

rasmith committed 11 months ago

[responsesAPI][6] Fix multi turn MCP tokenization (#30230)

Andrew Xia committed 11 months ago

[Attention] Make seq_lens_cpu optional in CommonAttentionMetadata to enable true async spec-decode (#29624)

Lucas Wilkinson committed 11 months ago

[Bugfix] Fix fp8 DeepGemm compilation issues (#30336)

ElizaWszola committed 11 months ago

[Bugfix] Fix cuda graph sizes when running with speculative decoding (#30330)

PatrykSaffer committed 11 months ago

[V1][Spec Decode] Optimize Medusa proposer to avoid GPU-CPU sync (#29723)

dongbo910220 committed 11 months ago

Improve wvsplitK tile and balance heristics. (#29937)

Hashem Hashemi committed 11 months ago

[Rocm][torch.compile] Adding layernorm + fp8 block quant and silu + fp8 block quant for Aiter (#25693)

Charlie Fu committed 11 months ago

[Quantization] FP8 Weight Reloading for Quantized RL Rollout (#28480)

Kyle Sayers committed 11 months ago

[MoE][Refactor] Remove most arguments to FusedMoEMethodBase.apply (#29066)

bnellnm committed 11 months ago

[CI/Build] Make test_mha_attn.py run on correct platform only and check for flash_attn_varlen_func in layer.py (#29145)

rasmith committed 11 months ago

Bump actions/stale from 10.1.0 to 10.1.1 (#30234)

dependabot[bot] committed 11 months ago

Bump actions/checkout from 6.0.0 to 6.0.1 (#30233)

dependabot[bot] committed 11 months ago

[Model][Quantization] Fix / Add GGUF support for Qwen2 MoE models (#30307)

Tsukasa OI committed 11 months ago

[CI/Test] Fix FP8 per-tensor quant test reference scale shape (#30352)

Lucas Wilkinson committed 11 months ago

[BugFix] Fix DeepSeek-R1 hang with DP and MTP (#30119)

Lucas Wilkinson committed 11 months ago