Making functions in utils/metrics.py compatible with torch.Tensor inputs
#216
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
…SNR` * updated tests to reflect changes
federico-carrara
requested review from
ashesh-0,
jdeschamps,
CatEek,
veegalinova and
melisande-c
…careamics into fc/refac/torch_metrics
ashesh-0
approved these changes
Aug 23, 2024
| Float tensor. | ||
| """ | ||
| if isinstance(x, torch.Tensor) and x.dtype != torch.float64: | ||
| warn( |
There was a problem hiding this comment.
Is the warning necessary? I ask because it will never lead to any information loss. Also, we do not return this tensor; we just use it to compute the metrics.
There was a problem hiding this comment.
I'd definitely remove it "in production", but I have temporarily put it there for debugging :)
| Predicted image. | ||
|
|
||
| Returns | ||
| ------- | ||
| Union[float, torch.tensor] | ||
| Scale invariant PSNR value. | ||
| """ | ||
| range_parameter = (np.max(gt) - np.min(gt)) / np.std(gt) | ||
| gt_ = _zero_mean(gt) / np.std(gt) | ||
| # cast tensors to double dtype |
4 tasks
|
What is the status of this PR? |
|
I remember there was a problem with dtypes changing from |
|
Let's leave it for now! |
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Outdated suggestions cannot be applied.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.
Description
This PR aims to make all the functions in
utils/metric.pytorch-compatible.torch.Tensorinputs.torch.Tensor's. Therefore, havingtorchcompatible metrics readily available avoids cumbersome type conversion within the NN models.numpy(e.g.,np.max(arr)) with more flexible method calls (e.g.,arr.max()).Changes Made
torchrequires tensors to be of typedouble(i.e.,torch.float64) to compute things like mean and std. Therefore, I added a private method that caststorch.tensor's todoubleto enablemean()andstd()computation. Casting tonumpyhappens instead in the PSNR computation that usesskimagemodule, since the module explicitly requires that. Finally, added notebook to show current issues with the tests.Please ensure your PR meets the following requirements: