torch estimator: check all ranks have same device type #2942
Conversation
This file contains 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
chongxiaoc
force-pushed
the
check_same_device
branch
from
8efcb87
to
10b9493
Compare
chongxiaoc
changed the title
tgaddair
reviewed
May 28, 2021
tgaddair
reviewed
May 28, 2021
chongxiaoc
force-pushed
the
check_same_device
branch
2 times, most recently
from
0bfaefa
to
9576eaa
Compare
tgaddair
reviewed
May 29, 2021
horovod/spark/lightning/remote.py
Outdated
| # We need to check all ranks have same device type for traning. | ||
| # Horovod doesn't support heterogeneous allreduce for gradients. | ||
| cuda_avail_list = hvd.allgather_object(cuda_available, name='device type') | ||
| assert cuda_avail_list.count(cuda_available) == hvd.size(), "All ranks don't have same device type!" |
There was a problem hiding this comment.
One thing I forgot to point out: this should be an exception (probably RuntimeError) rather than an assertion. Two reasons for this:
- Exceptions can be caught, whereas assertions effectively terminate the program, so there's nothing the caller can do to recover.
- Assertions can be removed by the interpreter at runtime when optimizers are enabled.
Assertions are primarily for debugging application code, whereas Exceptions are for errors at runtime.
There was a problem hiding this comment.
okay, I was thinking about raise assertion error instead of RuntimeError before.
Limitation: currently it cannot be applied to keras estimator, since allgather_object() requires tensorflow session is ready at that time. Signed-off-by: Chongxiao Cao <chongxiaoc@uber.com>
chongxiaoc
force-pushed
the
check_same_device
branch
from
9576eaa
to
a9364c8
Compare
|
@tgaddair Can you take one more look? |
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.
Signed-off-by: Chongxiao Cao chongxiaoc@uber.com
Checklist before submitting
Description
Fixes # (issue).
Review process to land