Additional error checking for torch.cuda.nccl APIs.
#43247
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
`torch.cuda.nccl` APIs didn't throw appropriate errors when called with inputs/outputs that were of the wrong type and it resulted in some cryptic errors instead. Adding some error checks with explicit error messages for these APIs. Differential Revision: [D23206069](https://our.internmc.facebook.com/intern/diff/D23206069/) [ghstack-poisoned]
pritamdamania87
pushed a commit
that referenced
this pull request
Aug 19, 2020
`torch.cuda.nccl` APIs didn't throw appropriate errors when called with inputs/outputs that were of the wrong type and it resulted in some cryptic errors instead. Adding some error checks with explicit error messages for these APIs. Differential Revision: [D23206069](https://our.internmc.facebook.com/intern/diff/D23206069/) ghstack-source-id: 110214324 Pull Request resolved: #43247
💊 CI failures summary and remediationsAs of commit e2f1eff (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group. This comment has been revised 10 times. |
mrshenli
approved these changes
Aug 19, 2020
torch/cuda/nccl.py
Outdated
| def _check_sequence_type(inputs): | ||
| if not isinstance(inputs, list) and not isinstance(inputs, tuple): | ||
| raise TypeError("inputs should be a list/tuple") | ||
|
|
rohan-varma
reviewed
Aug 19, 2020
torch/cuda/nccl.py
Outdated
|
|
||
| def _check_sequence_type(inputs): | ||
| if not isinstance(inputs, list) and not isinstance(inputs, tuple): | ||
| raise TypeError("inputs should be a list/tuple") |
Contributor
There was a problem hiding this comment.
Can it also be any python iterable/list-like object? I just tried the NCCL test with inputs = iter(inputs) (which makes it a list_iterable type) and inputs = set(inputs), and the current test_all_gather seems to pass. With this PR, would we stop supporting those types?
Contributor
Author
There was a problem hiding this comment.
I actually tried this approach out first, but apparently a standalone tensor is Iterable as well:
>>> isinstance(torch.rand(10), collections.Iterable)
True
As a result, if we check for just Iterable we might include a single Tensor as a valid input. I'll actually update the check with a combination of both checks to incorporate this.
`torch.cuda.nccl` APIs didn't throw appropriate errors when called with inputs/outputs that were of the wrong type and it resulted in some cryptic errors instead. Adding some error checks with explicit error messages for these APIs. Differential Revision: [D23206069](https://our.internmc.facebook.com/intern/diff/D23206069/) [ghstack-poisoned]
pritamdamania87
pushed a commit
that referenced
this pull request
Aug 21, 2020
Pull Request resolved: #43247 `torch.cuda.nccl` APIs didn't throw appropriate errors when called with inputs/outputs that were of the wrong type and it resulted in some cryptic errors instead. Adding some error checks with explicit error messages for these APIs. ghstack-source-id: 110484323 Differential Revision: [D23206069](https://our.internmc.facebook.com/intern/diff/D23206069/)
`torch.cuda.nccl` APIs didn't throw appropriate errors when called with inputs/outputs that were of the wrong type and it resulted in some cryptic errors instead. Adding some error checks with explicit error messages for these APIs. Differential Revision: [D23206069](https://our.internmc.facebook.com/intern/diff/D23206069/) [ghstack-poisoned]
pritamdamania87
pushed a commit
that referenced
this pull request
Aug 25, 2020
Pull Request resolved: #43247 `torch.cuda.nccl` APIs didn't throw appropriate errors when called with inputs/outputs that were of the wrong type and it resulted in some cryptic errors instead. Adding some error checks with explicit error messages for these APIs. ghstack-source-id: 110683546 Differential Revision: [D23206069](https://our.internmc.facebook.com/intern/diff/D23206069/)
Contributor
|
This pull request has been merged in 306eb3d. |
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 from ghstack:
torch.cuda.ncclAPIs. #43247 Additional error checking fortorch.cuda.ncclAPIs.torch.cuda.ncclAPIs didn't throw appropriate errors when calledwith inputs/outputs that were of the wrong type and it resulted in some cryptic
errors instead.
Adding some error checks with explicit error messages for these APIs.
Differential Revision: D23206069