[c10d] Fix data corruption bug after CUDAEventCache is enabled #138040
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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138040
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 31b6143 with merge base a77bb85 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
oncall: distributed
kwen2501
approved these changes
Oct 16, 2024
kwen2501
changed the title
kwen2501
changed the title
Here is why we see using `CUDAEventCache` cause crash and data corruption. 1. The deleter is doing its job and append the job the stack. 2. In create, instead of getting a reference, we are getting a copy of eventsArray_[i] (which is a std::vector). This is bad because we didn't really remove the element from the stack. While we thought we already pop up the last one from the stack, but it turns out the last one is still in the stack; we end up reusing the same event again and again. What's worse, since we keep adding new events to the stack, this will eventually explode the stack and a crash happens. Fix is easy, just get a reference. Local torchtitan run see a non-Nan loss. Also we want to use a deque instead of a stack, and refactor the code a bit to make it more readable. cc XilunWu H-Huang awgu kwen2501 wanchaol fegin wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
fduwjj
changed the title
wconstab
reviewed
Oct 16, 2024
wconstab
reviewed
Oct 16, 2024
shuqiangzhang
approved these changes
Oct 16, 2024
…bled" Here is why we see using `CUDAEventCache` cause crash and data corruption. 1. The deleter is doing its job and append the job the stack. 2. In create, instead of getting a reference, we are getting a copy of eventsArray_[i] (which is a std::vector). This is bad because we didn't really remove the element from the stack. While we thought we already pop up the last one from the stack, but it turns out the last one is still in the stack; we end up reusing the same event again and again. What's worse, since we keep adding new events to the stack, this will eventually explode the stack and a crash happens. Fix is easy, just get a reference. Local torchtitan run see a non-Nan loss. Also we want to use a deque instead of a stack, and refactor the code a bit to make it more readable. (in a separate PR) cc XilunWu H-Huang awgu kwen2501 wanchaol fegin wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
|
Separate fix and refactor so that it is better for reviewers. |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
|
You should probably add a unit test for the cache, that would have caught this. |
|
What I meant by unit test is literally a unit test of the cache. Put 3 events in, then get them back out and make sure you get the expected events and the cache is empty at the end. |
|
@wconstab Ahh, I see. Sure, let me write one. |
pytorchmergebot
pushed a commit
that referenced
this pull request
Oct 16, 2024
…138048) We used a LIFO stack to store the CudaEvent in the cache. ,Somehow we like FIFO deque better so aside from improving the readability of the code, we use a deque instead. As @wconstab pointed out, both methods are equally correct because the moment we put the event into stack/deque, the event is already ready for reuse, this change mostly is a preference change not trying to fix anything. Pull Request resolved: #138048 Approved by: https://github.com/kwen2501 ghstack dependencies: #138040
pytorchmergebot
pushed a commit
that referenced
this pull request
Oct 16, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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 (oldest at bottom):
Here is why we see using
CUDAEventCachecause crash and data corruption.Fix is easy, just get a reference. Local torchtitan run see a non-Nan loss.
Also we want to use a deque instead of a stack, and refactor the code a bit to make it more readable. (in a separate PR)
cc @XilunWu @H-Huang @awgu @kwen2501 @wanchaol @fegin @wz337 @wconstab @d4l3k @c-p-i-o