fix: for CUDA 11.1, caffe default on our Cudnn MIN_MEMORY handle #78
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
fantes
approved these changes
Oct 22, 2020
6 tasks
beniz
changed the title
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Outdated suggestions cannot be applied.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.
Other CUDNN handles with Caffe lead to unit tests failing, even with
batch_sizeset to 1. There must be some interaction with CUDNN and CUDA 11.1 that has become the norm, and that uses much more memory with other handle configurations.For this reason this PR forces MIN_MEMORY as the default, and I believe this is the right thing to do in all cases, and I've started defaulting to it a while ago through the API. The main reason is we want training/inference to have the highest changes of starting in most cases, then other optimizations can be tried such as modifying the CUDNN scheme to squeeze a bit more performances while using more memory.