Improving nnUNet inference speed #2048
Merged
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
* changing `lru_cache`'s `maxize=2` to `maxsize=None` for faster access and because cache will not be filled with more than 2 values
FabianIsensee
requested changes
Apr 9, 2024
* Added n_predictions back to replicate the previous behavior * Setting gaussian to 1 if not using the gaussian * setting lru cache size back to 2 to prevent OOM for unintended usage
|
Thanks! Had to make two small changes to make sure everything works + yields the same results. |
|
Tensors created during inference mode are slimmer because they do not have a version counter and requires_grad can't be set to True anymore. More information here: https://stackoverflow.com/a/74197846/18441695. |
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Outdated suggestions cannot be applied.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.
Main additions are:
torch.backends.cudnn.benchmark = True(used during training, but forgot during inference).torch.no_gradcontext with a singletorch.inference_modecontext.