Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 22, 2023

• docs: add finetuner to ppl who use accelerate (huggingface#1224)

  bwanglzu

  authored
  4eaf36e

  Copy full SHA for 4eaf36e

• Fix get_logger kwarg documentation issue (huggingface#1222)

  bcol23

  authored
  e7bb060

  Copy full SHA for e7bb060

Commits on Mar 20, 2023

• Fix bug in loading launch config (huggingface#1218)

  Show description for a15d307
  neumyor

  authored
  a15d307

  Copy full SHA for a15d307

• FIx TPU gradient state (huggingface#1219)

  muellerzr

  authored
  7e7f344

  Copy full SHA for 7e7f344

• ds offload optim fix to use CPUAdam (huggingface#1208)

  Show description for 10c6746
  pacman100

  authored
  10c6746

  Copy full SHA for 10c6746

Commits on Mar 19, 2023

• Fix example in accumulate method (huggingface#1211)

  VikParuchuri

  authored
  82c2665

  Copy full SHA for 82c2665

Commits on Mar 18, 2023

• Fix typo in TPU config (huggingface#1202)

  muellerzr

  authored
  2930cac

  Copy full SHA for 2930cac

Commits on Mar 16, 2023

• Better error message when using multi-GPU and Accelerate on torch <1.9.1 (huggingface#1203)

  Show description for 901ab69
  muellerzr

  authored
  901ab69

  Copy full SHA for 901ab69

• Fix tied weights load (huggingface#1204)

  Show description for 780e4aa
  sgugger

  authored
  780e4aa

  Copy full SHA for 780e4aa

Commits on Mar 15, 2023

• Make the Scheduler adjust the steps taken relative to the gradient accumulation steps (huggingface#1187)

  Show description for e462098
  muellerzr

  authored
  e462098

  Copy full SHA for e462098

• Fixup --fsdp (huggingface#1198)

  muellerzr

  authored
  017a98c

  Copy full SHA for 017a98c

• [Accelerator] We should not call to on modules that wraps accelerate loaded models (huggingface#1172)

  Show description for d1aa558
  younesbelkada

  authored
  d1aa558

  Copy full SHA for d1aa558

Commits on Mar 14, 2023

• Set drop last to ensure modulo16 restriction for fp8 (huggingface#1189)

  Show description for 41479fe
  ksivaman

  authored
  41479fe

  Copy full SHA for 41479fe

Commits on Mar 13, 2023

• Only convert linear layers with weights multiple of 16 (huggingface#1188)

  Show description for eac5d13
  sgugger

  authored
  eac5d13

  Copy full SHA for eac5d13

• add use_orig_params to FullyShardedDataParallelPlugin (huggingface#1184)

  Show description for b228136
  pacman100

  authored
  b228136

  Copy full SHA for b228136

• Add documentation about PyTorch FSDP state dict behavior (huggingface#1181)

  VikParuchuri

  authored
  90deb74

  Copy full SHA for 90deb74

• Support special mapping of dtypes when preparing device map (huggingface#1179)

  sgugger

  authored
  d942708

  Copy full SHA for d942708

Commits on Mar 10, 2023

• fixed typo in launch.py tpu_pod_launcher (huggingface#1180)

  hackpert

  authored
  3783180

  Copy full SHA for 3783180

• Add repr to AlignHook for easier debugging. (huggingface#1177)

  sgugger

  authored
  ea836f3

  Copy full SHA for ea836f3

• Run accelerate_test in cli (huggingface#1176)

  Show description for a4c9476
  muellerzr

  authored
  a4c9476

  Copy full SHA for a4c9476

• Fix CPU error always being raised (huggingface#1175)

  Show description for 3ca8c9a
  muellerzr

  authored
  3ca8c9a

  Copy full SHA for 3ca8c9a

Commits on Mar 9, 2023

• Fix accelerate test with new config_file errors (huggingface#1169)

  sgugger

  authored
  2f83b1a

  Copy full SHA for 2f83b1a

• Fix backward compatibility in configs wrt dynamo backend (huggingface#1168)

  sgugger

  authored
  b0591c6

  Copy full SHA for b0591c6

• v0.18.0.dev0

  sgugger

  committed
  d9871c0

  Copy full SHA for d9871c0

Commits on Mar 8, 2023

• Remove outdated command directions and use in tests (huggingface#1166)

  Show description for abc2beb
  muellerzr

  authored
  abc2beb

  Copy full SHA for abc2beb

• Fix what files get deleted through total_limit (huggingface#1165)

  Show description for 8749b4e
  muellerzr

  authored
  8749b4e

  Copy full SHA for 8749b4e

• Document skip_first_batches in the checkpoint usage guides (huggingface#1164)

  Show description for 4a3eaee
  muellerzr

  authored
  4a3eaee

  Copy full SHA for 4a3eaee

• [Accelerator] Fix issue with 8bit models (huggingface#1155)

  Show description for 3533e2b
  

  younesbelkada

  and

  sgugger

  authored
  3533e2b

  Copy full SHA for 3533e2b

• Attempt to fix import error when PyTorch is build without torch.distributed module (huggingface#1108)

  Show description for 3e0ceac
  mfuntowicz

  authored
  3e0ceac

  Copy full SHA for 3e0ceac

Commits on Mar 7, 2023

• Let GradientState know active dataloaders and reset the remainder (huggingface#1162)

  muellerzr

  authored
  03b617b

  Copy full SHA for 03b617b

• update support for torch dynamo compile (huggingface#1150)

  Show description for 840bb1a
  

  pacman100

  and

  sgugger

  authored
  840bb1a

  Copy full SHA for 840bb1a

• Fp8 integration (huggingface#1086)

  Show description for 1bfde6b
  

  sgugger

  and

  pacman100

  authored
  1bfde6b

  Copy full SHA for 1bfde6b

• 📝 add a couple more trackers to the docs (huggingface#1158)

  nateraw

  authored
  3482495

  Copy full SHA for 3482495

Commits on Mar 6, 2023

• Load custom state to cpu (huggingface#1156)

  Show description for 947b2a8
  Guangxuan-Xiao

  authored
  947b2a8

  Copy full SHA for 947b2a8

• Solve arrow keys being environment dependant for accelerate config

  p1atdev

  authored
  cac1ed4

  Copy full SHA for cac1ed4

Pagination

Previous

Next