Remove lazy behavior from MelScale #1636
Conversation
mthrok
left a comment
mthrok
left a comment
There was a problem hiding this comment.
Looks good. Left minor comments on test improvements.
I know these are not directly related to the PR, but otherwise we wouldn't be improving the tests.
| def test_batch_MelScale(self): | ||
| specgram = torch.randn(2, 31, 2786) | ||
| n_stft = 31 | ||
| specgram = torch.randn(2, n_stft, 2786) |
There was a problem hiding this comment.
Can you use the default value of n_stft=201, so that we can catch an error for default value first if that happens in the future?
Also not sure where the 2786 is coming from, but I think we can use something shorter, like 256.
|
|
||
| # Single then transform then batch | ||
| expected = torchaudio.transforms.MelScale()(specgram).repeat(3, 1, 1, 1) | ||
| expected = torchaudio.transforms.MelScale(n_stft=n_stft)(specgram).repeat(3, 1, 1, 1) |
There was a problem hiding this comment.
Note: (follow-up, not the scope of this PR) For batch consistency test, use of repeated does not catch some edge case. #1315
We should update the logic at some point.
| def test_melscale_load_save(self): | ||
| specgram = torch.ones(1, 1000, 100) | ||
| melscale_transform = transforms.MelScale() | ||
| melscale_transform = transforms.MelScale(n_stft=1000) |
There was a problem hiding this comment.
Instead of setting n_stft, can we change the shape of specgram to 1, 201, 100?
This is more of a personal preference but 1000 is somehow conspicuous.
Rebases pytorch#1571; addresses pytorch#1569: "In 0.9.0 we are deprecating the lazy behavior of MelScale because it can make an invalid TorchScript object and it does not align with the design of torchaudio. Now in master branch, we can remove the implementation." Co-authored-by: Pankaj Patil <[email protected]> Co-authored-by: moto <[email protected]> Co-authored-by: hwangjeff <[email protected]>
4 participants
Rebases #1571; addresses #1569:
"In 0.9.0 we are deprecating the lazy behavior of MelScale because it can make an invalid TorchScript object and it does not align with the design of torchaudio. Now in master branch, we can remove the implementation."