Commits

Branch selector

 master

User selector

All users

All time

Commit history

Commits on Oct 17, 2024

• host timer check from 2.5 to 2.6

  YizhouZ

  committed
  22e4f9c

  Copy full SHA for 22e4f9c

Commits on Oct 15, 2024

• Add README Pipeline Status for Huawei Ascend NPU (#6588)

  Show description for c9899dc
  

  5 peopleauthored
  c9899dc

  Copy full SHA for c9899dc

• Lock cache file of HF model list (#6628)

  Show description for 1a45bd8
  tohtana

  authored
  1a45bd8

  Copy full SHA for 1a45bd8

• add option to disable logger while compiling to avoid graph breaks (#6496)

  Show description for ce468c3
  

  3 peopleauthored
  ce468c3

  Copy full SHA for ce468c3

• Support safetensors export (#6579)

  Show description for bf60fc0
  

  3 peopleauthored
  bf60fc0

  Copy full SHA for bf60fc0

Commits on Oct 14, 2024

• Add first Step in LR Schedulers (#6597)

  Show description for 85b7469
  

  3 peopleauthored
  85b7469

  Copy full SHA for 85b7469

• Accept btl_tcp_if_include option through launcher_args (#6613)

  Show description for 13c16c9
  

  3 peopleauthored
  13c16c9

  Copy full SHA for 13c16c9

• Add API for updating ZeRO gradients (#6590)

  tjruwase

  authored
  65ab644

  Copy full SHA for 65ab644

• [compile] Show breakdown of graph break (#6601)

  Show description for cf41e8c
  delock

  authored
  cf41e8c

  Copy full SHA for cf41e8c

• Ignore reuse_dist_env (#6623)

  Show description for 7a5bc4f
  tohtana

  authored
  7a5bc4f

  Copy full SHA for 7a5bc4f

Commits on Oct 11, 2024

• apply fp16 autocast only to floating point values

  

  Masahiro Tanaka

  committed
  5c4b97f

  Copy full SHA for 5c4b97f

Commits on Oct 10, 2024

• Add API to get devices of offload states (#6586)

  Show description for adec991
  

  3 peopleauthored
  adec991

  Copy full SHA for adec991

• reduce setting global variables to reduce torch compile graph breaks (#6541)

  Show description for d7ca3d8
  

  4 peopleauthored
  d7ca3d8

  Copy full SHA for d7ca3d8

Commits on Oct 9, 2024

• AIO CPU Locked Tensor (#6592)

  Show description for a1f98bd
  

  jomayeri

  and

  tjruwase

  authored
  a1f98bd

  Copy full SHA for a1f98bd

• Clean up prefetched parameters (#6557)

  Show description for 7d751ee
  

  3 peopleauthored
  7d751ee

  Copy full SHA for 7d751ee

• Update version.txt after 0.15.2 release (#6615)

  Show description for 55f7f37
  

  loadams

  and

  jomayeri

  authored
  55f7f37

  Copy full SHA for 55f7f37

• Enabled Qwen2-MoE Tensor Parallelism (TP) inference (#6551)

  Show description for 474a328
  

  gyou2021

  and

  loadams

  authored
  474a328

  Copy full SHA for 474a328

• Unpin accelerate tests, update lightning with node16 removal. (#6611)

  Show description for 1062a0c
  loadams

  authored
  1062a0c

  Copy full SHA for 1062a0c

• Rearrange inference OPS and stop using builder.load (#5490)

  Show description for 645639b
  

  5 peopleauthored
  645639b

  Copy full SHA for 645639b

Commits on Oct 8, 2024

• Handle when `backend` is also in compile_kwargs (#6502)

  Show description for ca8b1fe
  

  4 peopleauthored
  ca8b1fe

  Copy full SHA for ca8b1fe

• Fix device selection using CUDA_VISIBLE_DEVICES (#6530)

  Show description for 5cbbff4
  

  3 peopleauthored
  5cbbff4

  Copy full SHA for 5cbbff4

• Improve DS logging control (#6602)

  Show description for f74ea69
  tjruwase

  authored
  f74ea69

  Copy full SHA for f74ea69

• Add llama3.2 vision autotp (#6577)

  Show description for e97b453
  

  Yejing-Lai

  and

  loadams

  authored
  e97b453

  Copy full SHA for e97b453

• Pin accelerate to fix CI failures/issues (#6610)

  loadams

  authored
  745dd48

  Copy full SHA for 745dd48

• Fix SD workflow (#6609)

  Show description for 00c4b98
  loadams

  authored
  00c4b98

  Copy full SHA for 00c4b98

• Move V100 workflows from cuda 11.1/11.7 to 12.1 (#6607)

  loadams

  authored
  20695b3

  Copy full SHA for 20695b3

Commits on Oct 7, 2024

• Add SSF Best practices badge (#6604)

  Show description for 940887d
  loadams

  authored
  940887d

  Copy full SHA for 940887d

• Cleanup CODEOWNERS file to be valid (#6603)

  loadams

  authored
  239b83a

  Copy full SHA for 239b83a

Commits on Oct 4, 2024

• [ROCm] Fix subprocess error (#6587)

  Show description for b93c7a2
  jagadish-amd

  authored
  b93c7a2

  Copy full SHA for b93c7a2

Commits on Sep 27, 2024

• Fix torch include in `op_builder/mlu/fused_adam.py` and update no-torch workflow triggers (#6584)

  Show description for 8cded57
  loadams

  authored
  8cded57

  Copy full SHA for 8cded57

• Fixes on the accelerate side mean we do not need to skip this test (#6583)

  Show description for 828ddfb
  loadams

  authored
  828ddfb

  Copy full SHA for 828ddfb

• [COMPILE] workflow for deepspeed + torch.compile (#6570)

  Show description for d4e1895
  

  YizhouZ

  and

  tohtana

  authored
  d4e1895

  Copy full SHA for d4e1895

• add bfloat16 to inference support dtypes (#6528)

  Show description for 1caf6e8
  

  3 peopleauthored
  1caf6e8

  Copy full SHA for 1caf6e8

• Add APIs to offload states of model, optimizer, and engine (#6011)

  Show description for 047bcf6
  

  3 peopleauthored
  047bcf6

  Copy full SHA for 047bcf6

Commits on Sep 26, 2024

• [XPU] Support DeepNVMe new code structure (#6532)

  Show description for d45cfd3
  

  3 peopleauthored
  d45cfd3

  Copy full SHA for d45cfd3