Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

Commits on Sep 5, 2025

• Use Pytorch's AMP for amp test (#1929)

  tinglvv

  authored
  c54dbd3

  Copy full SHA for c54dbd3

Commits on Jul 16, 2021

• local_rank fix (#1129)

  Show description for c1378e6
  xwang233

  authored
  c1378e6

  Copy full SHA for c1378e6

Commits on Dec 2, 2020

• Fix lack of proper loading of best_prec1 from the checkpoint (#1000)

  Show description for 6c186b3
  JanuszL

  authored
  6c186b3

  Copy full SHA for 6c186b3

Commits on Sep 29, 2020

• use reshape instead of view (#971)

  ptrblck

  authored
  a109f85

  Copy full SHA for a109f85

Commits on Jan 27, 2020

• Channels last support (#668)

  

  VitalyFedyunin

  authored and

  mcarilli

  committed
  2ca894d

  Copy full SHA for 2ca894d

Commits on Oct 22, 2019

• Making the encouragement to use O1 a bit stronger...

  definitelynotmcarilli

  committed
  95d6c00

  Copy full SHA for 95d6c00

Commits on Jul 27, 2019

• Put parser in a function to make script importable

  syed-ahmed

  committed
  e86f986

  Copy full SHA for e86f986

Commits on Jun 21, 2019

• Make main_amp.py more profiling-friendly

  definitelynotmcarilli

  committed
  f29b3f8

  Copy full SHA for f29b3f8

Commits on Jun 20, 2019

• Fix end-of-epoch with record_stream

  definitelynotmcarilli

  committed
  90e5b05

  Copy full SHA for 90e5b05

Commits on Jun 18, 2019

• Fix rare caching allocator race condition in imagenet prefetcher

  definitelynotmcarilli

  committed
  d5e2bb4

  Copy full SHA for d5e2bb4

Commits on Apr 30, 2019

• Remove unused tensor in fast_collate (#281)

  Show description for ca2baff
  

  ptrblck

  authored and

  mcarilli

  committed
  ca2baff

  Copy full SHA for ca2baff

Commits on Apr 29, 2019

• Warning for inception_v3

  definitelynotmcarilli

  committed
  7b245db

  Copy full SHA for 7b245db

Commits on Mar 14, 2019

• Adding simple distributed example for #200

  definitelynotmcarilli

  committed
  a730f38

  Copy full SHA for a730f38

Commits on Mar 13, 2019

• Casting model output as well as input, for #195

  definitelynotmcarilli

  committed
  d1f74a3

  Copy full SHA for d1f74a3

Commits on Mar 8, 2019

• Fix for #188

  definitelynotmcarilli

  committed
  a3a09c8

  Copy full SHA for a3a09c8

Commits on Mar 7, 2019

• Update README.md

  

  mcarilli

  authored
  40555b3

  Copy full SHA for 40555b3

• Update README.md

  

  mcarilli

  authored
  a74a0f2

  Copy full SHA for a74a0f2

• Update README.md

  

  mcarilli

  authored
  d2862d5

  Copy full SHA for d2862d5

Commits on Mar 4, 2019

• Cleaning up READMEs

  definitelynotmcarilli

  committed
  df83b67

  Copy full SHA for df83b67

• Minor typos

  definitelynotmcarilli

  committed
  b90b057

  Copy full SHA for b90b057

• Adding summary

  definitelynotmcarilli

  committed
  a3dbea3

  Copy full SHA for a3dbea3

• README touchup

  definitelynotmcarilli

  committed
  26b30d1

  Copy full SHA for 26b30d1

• Comprehensive imagenet example for the new API

  definitelynotmcarilli

  committed
  254cad2

  Copy full SHA for 254cad2

Commits on Feb 25, 2019

• Fix for unscale usage in fp16_utils.FP16_Optimizer

  definitelynotmcarilli

  committed
  ed8236f

  Copy full SHA for ed8236f

Commits on Feb 19, 2019

• Reworked multi tensor apply, added tests

  definitelynotmcarilli

  committed
  6763a8b

  Copy full SHA for 6763a8b

Commits on Feb 1, 2019

• async->non_blocking, module-specific logging

  definitelynotmcarilli

  committed
  cc85a2e

  Copy full SHA for cc85a2e

• Making static loss scale the default, and clipping master grads when running with --fp16

  definitelynotmcarilli

  committed
  43522e6

  Copy full SHA for 43522e6

Commits on Jan 5, 2019

• Adding amp imagenet example

  definitelynotmcarilli

  committed
  3c7a0e4

  Copy full SHA for 3c7a0e4

Commits on Dec 4, 2018

• call .float() on GPU, remove unnecessary push to GPU

  ptrblck

  committed
  9ccebe5

  Copy full SHA for 9ccebe5

Commits on Dec 2, 2018

• update examples to PyTorch >=0.4.0

  ptrblck

  committed
  28bdc04

  Copy full SHA for 28bdc04

Commits on Nov 30, 2018

• Syncing main.py a bit more

  definitelynotmcarilli

  committed
  bc62f32

  Copy full SHA for bc62f32

• Adding deterministic option to main_fp16_optimizer.py

  definitelynotmcarilli

  committed
  2a8022c

  Copy full SHA for 2a8022c

Commits on Nov 10, 2018

• Updating example instructions to use batch size 224 for safety

  definitelynotmcarilli

  committed
  2b8277e

  Copy full SHA for 2b8277e

Commits on Oct 29, 2018

• Merging in fused adam optimizer, additional DDP features tested in 18.10 (#60)

  Show description for e0bc5d6
  

  mcarilli

  authored
  e0bc5d6

  Copy full SHA for e0bc5d6

Commits on Oct 23, 2018

• [syncBN] (#48)

  Show description for 81eef1e
  

  jjsjann123

  authored and

  mcarilli

  committed
  81eef1e

  Copy full SHA for 81eef1e