- 05 Aug, 2020 1 commit
-
-
ngimel authored
* add device guards to the optimizers * add untracked file * set deviceGuard in multi_tensor_apply * address review comments; fix lamb * indent * typo
-
- 21 May, 2020 1 commit
-
-
Jeff Daily authored
-
- 20 May, 2020 1 commit
-
-
lcskrishna authored
-
- 12 May, 2020 1 commit
-
-
rohithkrn authored
-
- 03 Jul, 2019 1 commit
-
-
Michael Carilli authored
-
- 27 May, 2019 1 commit
-
-
Michael Carilli authored
-
- 10 May, 2019 1 commit
-
-
Michael Carilli authored
-
- 26 Apr, 2019 2 commits
-
-
Michael Carilli authored
-
Michael Carilli authored
-
- 25 Apr, 2019 1 commit
-
-
Michael Carilli authored
-
- 22 Apr, 2019 1 commit
-
-
Michael Carilli authored
-
- 18 Apr, 2019 1 commit
-
-
Michael Carilli authored
-
- 11 Mar, 2019 2 commits
-
-
Simon Layton authored
-
Simon Layton authored
Fix dispatch where we have a parameter group with multiple combinations of types Optionally apply weight decay after momentum
-
- 09 Mar, 2019 1 commit
-
-
Simon Layton authored
-
- 08 Mar, 2019 5 commits
-
-
Simon Layton authored
-
Simon Layton authored
Incorrect types used in a few places
-
Simon Layton authored
Only support the 4 specific cases we care about Remove more general set of switch statements
-
Simon Layton authored
Fuse in fp16 gradient -> fp32 convert Additional option fp16 weight copy written out
-
Simon Layton authored
Initial implementation, all fp32 Tested against torch.optim.sgd
-
