[DeepEPv2] Support MXFp8 Activation Scale Dispatch (#51398)

Robert Shaw committed about 2 months ago

[Doc] Fix local input path in run-batch examples across docs (#53220)

QWERQWERQWE86 committed about 2 months ago

[MoE] enable all2all fi_one_sided by default (#53311)

Artem Perevedentsev committed about 2 months ago

[CI/Build] Improve pre-commit fail message (#53866)

Cyrus Leung committed about 2 months ago

[Offloader] Offload submodules that make_layers never reaches (#53120)

ray24777 committed about 2 months ago

[CI] forward fix CRCR report step in the torch-nightly lane (#53732)

Andrey Talman committed about 2 months ago

[CI] Build mamba-ssm with C++20 for torch 2.14 nightly compatibility (#49600)

Andrey Talman committed about 2 months ago

[Bugfix][Processor] Replace bare asserts with ValueError in DeepseekVLV2/OCR processors (#53854)

adisivaprasad committed about 2 months ago

[AttentionBackend][HPC-ops] update hpc rope norm to support stride kv cache (#53705)

Cheng Jiang committed about 2 months ago

[LoRA] Cleanup VocabParallelEmbedding (#53843)

Jee Jee Li committed about 2 months ago

[Config] Delegate PCP compatibility checks to PCP manager (#53853)

Qiu Chunshuo committed about 2 months ago

[Bugfix][Mooncake] Fix Mamba prefill truncation ordering (#53663)

Ziming Huang committed about 2 months ago

[CI/Build][The Rock] Use model_class_overrides so spawned worker can use test PredictableLlamaForCausalLM class when worker spawned using Python 3.14 (#49218)

rasmith committed about 2 months ago

[Hardware][AMD][Perf][Bugfix] Update ROCr and clr in base image (#53712)

Matt committed about 2 months ago

[RL] Add rank-local IPC weight updates (#52497)

aoshen02 committed about 2 months ago

[Bugfix][Model] Honor Molmo2 dummy video num_frames >= 2 override (#53830)

Hung Hoang (黄兴） committed about 2 months ago

BugFix(PCP): use persistent input buffers for PIECEWISE CUDA graphs (#53515)

Qiu Chunshuo committed about 2 months ago

[Bugfix][Structured Output] Preserve reasoning adapters for shared parser engines (#52830)

Flora Feng committed about 2 months ago

[Kernel][Perf] Tune fused_moe FP8 config for Qwen3.5 on L40S (+7%) (#53819)

nicholaskh-ai committed about 2 months ago

[Frontend] Add `/v1/messages/render` endpoint for the Anthropic Messages API (#45803)

Hyunkyun Moon committed about 2 months ago