Commits · 1579b9e3db87517e75933e08707b15e074717f2f · OpenDAS / apex

10 Sep, 2018 4 commits
- Merge pull request #40 from NVIDIA/amp_tests · 1579b9e3
  Carl Case authored Sep 10, 2018
```
amp unit tests
```
  1579b9e3
- add rnn tests · 5f5dfa42
  Carl Case authored Sep 10, 2018
  
  5f5dfa42
- tests on banned methods · 39928327
  Carl Case authored Sep 10, 2018
  
  39928327
- add more promotion testing · 22920fe0
  Carl Case authored Sep 10, 2018
  
  22920fe0
07 Sep, 2018 1 commit
- Adding verbose option to FP16_Optimizer · 75a865e3
  Michael Carilli authored Sep 07, 2018
  
  75a865e3
06 Sep, 2018 2 commits
- Enabling single-process fallback for examples/imagenet/main_reducer.py · cb6d8f1a
  Michael Carilli authored Sep 06, 2018
  
  cb6d8f1a
- Revising LR scaling to account for any choice of num processes, batch size per process · a2801d91
  Michael Carilli authored Sep 06, 2018
  
  a2801d91
05 Sep, 2018 2 commits
- minor fix · 01e29c97
  Michael Carilli authored Sep 05, 2018
  
  01e29c97
- Fixing needs_refresh logic to allow multiple forwards between each backward · ed14f39c
  Michael Carilli authored Sep 05, 2018
  
  ed14f39c
30 Aug, 2018 2 commits
- Update README.md · 586c507e
  mcarilli authored Aug 30, 2018
  
  586c507e
- Update distributed.py · 5a39c5e3
  mcarilli authored Aug 30, 2018
  
  5a39c5e3
28 Aug, 2018 6 commits
- Updating imagenet README · 034b8f02
  Michael Carilli authored Aug 28, 2018
  
  034b8f02
- Cleaning up git weirdness + updating docs for Reducer · 37a4b221
  Michael Carilli authored Aug 28, 2018
  
  37a4b221
- Reducer (#38) · f09fb4f4
  Christian Sarofeen authored Aug 28, 2018
```
* Add reducer class in parallel/distributed.

* Separate DDP and Reducer examples.

* Don't confuse DDP and reducer in example.
```
  f09fb4f4
- Add reducer class in parallel/distributed. (#37) · 73b62dde
  Christian Sarofeen authored Aug 28, 2018
  
  73b62dde
- Adjusting learning rate for batch size · dc41c5ce
  Michael Carilli authored Aug 27, 2018
  
  dc41c5ce
- WIP: promotion tests · 81c788f0
  Carl Case authored Aug 27, 2018
  
  81c788f0
27 Aug, 2018 2 commits
- add first basic amp unit tests · 2e69d933
  Carl Case authored Aug 27, 2018
  
  2e69d933
- experimental: ability to deactivate amp with handle · ebaa5a15
  Carl Case authored Aug 27, 2018
  
  ebaa5a15
22 Aug, 2018 1 commit
- Update README.md · 437bcf22
  mcarilli authored Aug 22, 2018
  
  437bcf22
20 Aug, 2018 2 commits
- Disabling import of experimental/planned files · 12dce88d
  Michael Carilli authored Aug 19, 2018
  
  12dce88d
- minor cleanup · 59bf7d13
  Michael Carilli authored Aug 20, 2018
  
  59bf7d13
19 Aug, 2018 2 commits
- updating FP16_Optimizer example as well · 17971202
  Michael Carilli authored Aug 18, 2018
  
  17971202
- Adjusting learning rate schedule for 76% accuracy · eae8b989
  Michael Carilli authored Aug 18, 2018
  
  eae8b989
16 Aug, 2018 3 commits
- Syncing imagenet examples to use distributed validation · 39c9be85
  Michael Carilli authored Aug 16, 2018
  
  39c9be85
- Merge branch 'dist_valid' · 7d0aef9f
  Michael Carilli authored Aug 16, 2018
  
  7d0aef9f
- Removing orphaned /distributed/run_distributed.sh · 2af29c19
  Michael Carilli authored Aug 16, 2018
  
  2af29c19
14 Aug, 2018 1 commit
- Fix to enable freezing params · 1d45fada
  Michael Carilli authored Aug 14, 2018
  
  1d45fada
07 Aug, 2018 1 commit
- Dist validation imagenet main.py · ae3be17a
  Christian Sarofeen authored Aug 07, 2018
  
  ae3be17a
25 Jul, 2018 1 commit
- Using temporary Python-side inf+nan check in amp/scaler.py · 2063287b
  Michael Carilli authored Jul 25, 2018
  
  2063287b
24 Jul, 2018 1 commit
- Fixing FP16_Optimizer handling of LBFGS · ae921de2
  Michael Carilli authored Jul 24, 2018
  
  ae921de2
23 Jul, 2018 2 commits
- Switch to simple Python-only install, in preparation for upstreaming C++ backend. · d695b68b
  Michael Carilli authored Jul 23, 2018
  
  d695b68b
- Ported examples to use torch.distributed.launch · aa817132
  Michael Carilli authored Jul 19, 2018
  
  aa817132
18 Jul, 2018 3 commits
- delete __getstate__ arg · 12b49b98
  ngimel authored Jul 18, 2018
  
  12b49b98
- Merge pull request #30 from NVIDIA/checkpoint_fix · 77ee4bcd
  Christian Sarofeen authored Jul 18, 2018
```
Handle set/get state for DDP, remove stream which cant be pickled.
```
  77ee4bcd
- Handle set/get state for DDP, remove stream which cant be pickled. · f1f97f9f
  Christian Sarofeen authored Jul 18, 2018
  
  f1f97f9f
09 Jul, 2018 2 commits
- FP16_Optimizer now preserves param order and casts per-param state tensors to FP32 · 5c6144e6
  Michael Carilli authored Jul 09, 2018
  
  5c6144e6
- Fix single-GPU fallback for examples/distributed · 4a8cf7ad
  mcarilli authored Jul 09, 2018
  
  4a8cf7ad
05 Jul, 2018 2 commits
- Update README with Windows build instructions · 31931985
  mcarilli authored Jul 05, 2018
  
  31931985
- Update README.md · 738f7607
  mcarilli authored Jul 05, 2018
  
  738f7607