Code Quality

Workflow runs for Code Quality

• Speed up CI JIT compiles: Triton cache and arch-limited CUDA wheels

  completed successfully:
  Code Quality#13891:Pull request#2905openedbyshengliangxu
  shengliangx/ci-shared-cache

  Oct 21, 20265m 57s
  More actions for workflow run 13891

• Code Quality

  completed successfully:
  Code Quality#13890:Scheduled
  main

  Oct 21, 20264m 40s
  More actions for workflow run 13890

• Fix attention quantizer leakage in diffusers HF exports

  completed successfully:
  Code Quality#13889:Pull request#2904synchronizebyZhehao-Hu
  fix/nvbug-6927340-qwen-attention-export

  Oct 20, 20265m 36s
  More actions for workflow run 13889

• megatron_bridge: load the HF model from local disk; Megatron export copies its non-model files

  completed successfully:
  Code Quality#13888:Pull request#2610synchronizebyshengliangxu
  shengliangx/model-download-4-megatron

  Oct 20, 20265m 35s
  More actions for workflow run 13888

• Fix attention quantizer leakage in diffusers HF exports

  completed successfully:
  Code Quality#13887:Pull request#2904openedbyZhehao-Hu
  fix/nvbug-6927340-qwen-attention-export

  Oct 20, 20266m 30s
  More actions for workflow run 13887

• Export GGML-quantized layers as INT8 compressed-tensors for vLLM

  completed successfully:
  Code Quality#13886:Pull request#2903synchronizebycjluo-nv
  chenjiel/ggml-int8-export

  Oct 20, 20265m 22s
  More actions for workflow run 13886

• megatron_bridge: load the HF model from local disk; Megatron export copies its non-model files

  completed successfully:
  Code Quality#13885:Pull request#2610synchronizebyshengliangxu
  shengliangx/model-download-4-megatron

  Oct 20, 20265m 53s
  More actions for workflow run 13885

• Export GGML-quantized layers as INT8 compressed-tensors for vLLM

  completed successfully:
  Code Quality#13884:Pull request#2903openedbycjluo-nv
  chenjiel/ggml-int8-export

  Oct 20, 20265m 57s
  More actions for workflow run 13884

• Fix vanilla DSpark checkpoint exports for vLLM

  completed successfully:
  Code Quality#13883:Pull request#2902openedbysragrawal
  sragrawal:sandeepa/dspark-vllm-export

  Oct 20, 20265m 53s
  More actions for workflow run 13883

• [OMNIML-6054] [5/21] Decide how a model is distributed, from the model built on the meta device

  completed successfully:
  Code Quality#13882:Pull request#2737synchronizebyshengliangxu
  shengliangx/hfep-2-plan-decision

  Oct 20, 20267m 35s
  More actions for workflow run 13882

• [OMNIML-6076] [13/21] Describe expert parallelism in the model specs

  completed successfully:
  Code Quality#13881:Pull request#2746synchronizebyshengliangxu
  shengliangx/hfep-9-ep-model-specs

  Oct 20, 20266m 17s
  More actions for workflow run 13881

• [OMNIML-6079] [20/21] Add the Kimi-K3 expert-parallel adapter

  completed successfully:
  Code Quality#13880:Pull request#2749synchronizebyshengliangxu
  shengliangx/hfep-11c-kimi-k3

  Oct 20, 20267m 25s
  More actions for workflow run 13880

• [OMNIML-6078] [19/21] Add the Kimi-K2.5 and Kimi-VL expert-parallel adapters

  completed successfully:
  Code Quality#13879:Pull request#2748synchronizebyshengliangxu
  shengliangx/hfep-11b-kimi-k25-vl

  Oct 20, 20265m 21s
  More actions for workflow run 13879

• [OMNIML-6085] [21/21] Run expert-parallel PTQ from hf_ptq.py

  completed successfully:
  Code Quality#13878:Pull request#2754synchronizebyshengliangxu
  shengliangx/hfep-14-hf-ptq

  Oct 20, 20266m 31s
  More actions for workflow run 13878

• [OMNIML-6078] [19/21] Add the Kimi-K2.5 and Kimi-VL expert-parallel adapters

  cancelled:
  Code Quality#13877:Pull request#2748synchronizebyshengliangxu
  shengliangx/hfep-11b-kimi-k25-vl

  Oct 20, 20262s
  More actions for workflow run 13877

• [OMNIML-6064] [7/21] Load each rank's checkpoint shard straight into a meta-built model

  completed successfully:
  Code Quality#13876:Pull request#2739synchronizebyshengliangxu
  shengliangx/hfep-4-meta-dcp-load

  Oct 20, 20268m 23s
  More actions for workflow run 13876

• [OMNIML-6054] [5/21] Decide how a model is distributed, from the model built on the meta device

  cancelled:
  Code Quality#13875:Pull request#2737synchronizebyshengliangxu
  shengliangx/hfep-2-plan-decision

  Oct 20, 20261m 35s
  More actions for workflow run 13875

• [OMNIML-6079] [20/21] Add the Kimi-K3 expert-parallel adapter

  cancelled:
  Code Quality#13874:Pull request#2749synchronizebyshengliangxu
  shengliangx/hfep-11c-kimi-k3

  Oct 20, 20261m 36s
  More actions for workflow run 13874

• [OMNIML-6082] [15/21] Add the pieces of a distributed HF checkpoint writer

  completed successfully:
  Code Quality#13873:Pull request#2751synchronizebyshengliangxu
  shengliangx/hfep-12b-writer-pieces

  Oct 20, 20264m 38s
  More actions for workflow run 13873

• [OMNIML-6059] [1/21] Add the DP=EP all-to-all expert transport

  completed successfully:
  Code Quality#13872:Pull request#2738synchronizebyshengliangxu
  shengliangx/hfep-3-a2a-transport

  Oct 20, 20267m 43s
  More actions for workflow run 13872

• [OMNIML-6072] [2/21] Make amax synchronization correct under data-parallel calibration

  completed successfully:
  Code Quality#13871:Pull request#2744synchronizebyshengliangxu
  shengliangx/hfep-7-amax-sync

  Oct 20, 20266m 44s
  More actions for workflow run 13871

• [OMNIML-6062] [11/21] Dequantize block-FP8 checkpoints as they load expert-parallel

  completed successfully:
  Code Quality#13870:Pull request#2743synchronizebyshengliangxu
  shengliangx/hfep-6c-fp8-dequant

  Oct 20, 20266m 4s
  More actions for workflow run 13870

• [OMNIML-6063] [18/21] Decompress compressed-tensors MoE experts on the fly as they load

  completed successfully:
  Code Quality#13869:Pull request#2747synchronizebyshengliangxu
  shengliangx/hfep-11a-onfly-decompress

  Oct 20, 20267m 5s
  More actions for workflow run 13869

• [OMNIML-6083] [16/21] Write a sharded model's HF checkpoint without gathering it

  completed successfully:
  Code Quality#13868:Pull request#2752synchronizebyshengliangxu
  shengliangx/hfep-12c-nogather-export

  Oct 20, 20264m 51s
  More actions for workflow run 13868

• [OMNIML-6071] [8/21] Calibrate a distributed model on per-rank data

  completed successfully:
  Code Quality#13867:Pull request#2740synchronizebyshengliangxu
  shengliangx/hfep-5-distributed-calibration

  Oct 20, 20267m 1s
  More actions for workflow run 13867