Allow for iterations where no module parameter is used #19821
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
pytorchbot
added
the
oncall: distributed
pietern
force-pushed
the
export-D15113272
branch
from
ee3722a to
398fc2d
Compare
Differential Revision: D15113272 Differential Version: 80846101
pietern
force-pushed
the
export-D15113272
branch
from
398fc2d to
4ee9dd5
Compare
Differential Revision: D15113272 Differential Version: 80846136
pietern
changed the title
mrshenli
reviewed
Apr 27, 2019
| this->finalize_backward(); | ||
| }); | ||
| } else { | ||
| finalize_backward(); |
Contributor
There was a problem hiding this comment.
I am debating with myself whether we should support this or error out with explicit message. What are the use cases for DDP without backward?
Contributor
There was a problem hiding this comment.
we should definitely support it. It's possible that for a few iterations there's no backward and for a few iterations there is backward (for example when doing MCTS / rollouts)
soumith
approved these changes
Apr 27, 2019
Contributor
|
This pull request has been merged in 9b69da2. |
zhangguanheng66
pushed a commit
to zhangguanheng66/pytorch
that referenced
this pull request
May 6, 2019
Summary: Pull Request resolved: pytorch#19821 It is possible that not a single parameter is used during an iteration. If this is the case, the `prepare_for_backward` function marks all parameters as unused, kicks off reduction of all buckets, *and* finalizes the reduction. This is different from the prior implementation where we assumed that autograd would produce a gradient for at least a single parameter. We then used the autograd callback mechanism to queue a finalizer callback. Now, this finalizer may be executed in line. Reviewed By: mrshenli Differential Revision: D15113272 fbshipit-source-id: dc91458b569cd8c106ddaeea558464b515683550
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack:
:black_circle: #19821 Allow for iterations where no module parameter is used 💛
It is possible that not a single parameter is used during an
iteration. If this is the case, the
prepare_for_backwardfunctionmarks all parameters as unused, kicks off reduction of all buckets,
and finalizes the reduction.
This is different from the prior implementation where we assumed that
autograd would produce a gradient for at least a single parameter.
We then used the autograd callback mechanism to queue a finalizer
callback. Now, this finalizer may be executed in line.
Differential Revision: D15113272