forked from facebookresearch/fairseq
-
Notifications
You must be signed in to change notification settings - Fork 7
Pull requests: pytorch-tpu/fairseq
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Bug fix; change
trainers in assert_on_losses function definiton to trainer
#14
by taylanbil
was closed Oct 28, 2019
Loading…
In MP, trainers do not exist any more, so assert using single trainer
#13
by taylanbil
was merged Oct 25, 2019
Loading…
None input for loss assertion should be filled to inf in order not to throw
#11
by taylanbil
was merged Oct 18, 2019
Loading…
Add option to assert on training and/or validation loss
#10
by taylanbil
was merged Oct 18, 2019
Loading…
[DO NOT MERGE][RFC] Adding tpu capabilities to train.py
#7
by taylanbil
was closed Oct 3, 2019
Loading…
Change masked_fill_ input in loss in order to accomodate necessary pytorch changes
#6
by taylanbil
was merged Oct 1, 2019
Loading…
isolated chkp path to a util function so I can use it in the runner
#3
by taylanbil
was merged Aug 30, 2019
Loading…
Previous Next
ProTip!
Type g i on any issue or pull request to go back to the issue listing page.