Use a CUDAGuard when running Torch models #340
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
|
|
||
| #ifndef __APPLE__ | ||
| // Make sure we're running on the correct device | ||
| std::unique_ptr<at::cuda::CUDAGuard> device_guard; |
| const auto model_device = get_torch_device(DeviceType::GPU); | ||
| if (model_device.is_cuda()) | ||
| { | ||
| device_guard = stdx::make_unique<at::cuda::CUDAGuard>(model_device); |
Contributor
There was a problem hiding this comment.
I guess we can use std:: here not stdx:: because Neuropod became C++14 recently, right?
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Outdated suggestions cannot be applied.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.
This PR ensures that we're running on the correct device even if something else calls
cudaSetDevicebefore running inference.This fixes a class of issues where another piece of code changes the current device for the current thread. For example, this can happen if TF and Torch run together on the same threadpool. TF will call
cudaSetDeviceand cause torch to break if it runs on the same thread in the future.This can cause some obscure cuDNN errors and generally hard-to-debug issues.