Use a ptr to store autograd profiler rng #24911
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
Trying to fix #2575. [Here](https://gist.github.com/suo/7b0bc4b49d3c9e095b9f7eef8fa7c6e8) is all TLS in libtorch.so (thanks @ezyang for figuring how to find this) I noticed that `CallbackManager::sample_zero_one()::gen` has size 5000, which seems a lot bigger than the other ones. So make it heap-allocated instead. Caveat: I have no idea if this will actually fix anything, or whether making this variable heap-allocated is a bad idea. [pytorch ci] [win ci] [caffe2 ci] [binary ci]
pytorchbot
added
the
module: autograd
suo
added a commit
that referenced
this pull request
Aug 20, 2019
Trying to fix #2575. [Here](https://gist.github.com/suo/7b0bc4b49d3c9e095b9f7eef8fa7c6e8) is all TLS in libtorch.so (thanks @ezyang for figuring how to find this) I noticed that `CallbackManager::sample_zero_one()::gen` has size 5000, which seems a lot bigger than the other ones. So make it heap-allocated instead. Caveat: I have no idea if this will actually fix anything, or whether making this variable heap-allocated is a bad idea. [pytorch ci] [win ci] [caffe2 ci] [binary ci] ghstack-source-id: 262d66a Pull Request resolved: #24911
zdevito
approved these changes
Aug 20, 2019
Contributor
|
You need to explicitly enable the 3.5 test to see if this worked. You do this by (1) marking this config as XImportant in |
Contributor
|
This is probably right and we should land it ASAP in any case. |
Contributor
|
We should also just switch to an RNG which has a significantly smaller state than this one |
Contributor
|
For the record, this PR does NOT actually work. More discussion at #2575 (comment) |
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:
Trying to fix #2575.
Here is
all TLS in libtorch.so (thanks @ezyang for figuring how to find this)
I noticed that
CallbackManager::sample_zero_one()::genhas size 5000,which seems a lot bigger than the other ones. So make it heap-allocated
instead.
Caveat: I have no idea if this will actually fix anything, or whether
making this variable heap-allocated is a bad idea.
[pytorch ci] [win ci] [caffe2 ci] [binary ci]
Differential Revision: D16936370