added check for N=0

[rafia17]

No description provided.

[alecgunny]

These looks great. Couple points of feedback:

• It might be helpful to have some tests for these two. For the N=0 issue, you can use unittest.mock.patch to make torch.rand return all 0s to ensure the N=0 case gets triggered and then do some tests on that output to ensure it's what you expect. For the length issue, it should be enough to use a pytest.raises context and call the function with bad values to make sure the appropriate error is getting raised.
• I think the pre-commit checks are failing because you have unnecessary empty lines in the code. As a rule, we only allow double empty lines between top-level class and function definitions, not inside them.

[alecgunny]

See my notes in the comment

[rafia17]

Thanks for the feedback:

• Regarding testing, I have tested both instances extensively in last few weeks. Printing out values for mask on every run and verifying that if N < 0: block only gets executed for the correct value of the mask and subsequently for N.
• Similarly tested for the psd_length such that the error gets raised only when psd_length is less than the window length, otherwise it doesn't get raised.
• Regarding pre-commit checks failing, I have checked yet again but cannot see any double empty lines in either augmentor.py or train.py. So I am very puzzled here.