Example code for Spectrogram in documentation

[harishsdev]

[harishsdev]

PR for : #1564

[mthrok]

Thanks for the con tribulation but your base branch is old more than one month. Please rebase it on master.

[mthrok]

1. Indentation is off.
2. hyphen and colon are unnecessary.

Please follow an existing example;

audio/torchaudio/functional/functional.py

Lines 821 to 824 in aec0e8c

	Example
	>>> specgram = torch.randn(1, 40, 1000)
	>>> delta = compute_deltas(specgram)
	>>> delta2 = compute_deltas(delta)

[mthrok]

This example spends 6 lines for loading audio tensor and 3 lines for the Transform. Instead please make an example around the transform.

Make the resulting transform instance an intermediate named variable.
Also please use some constructor arguments for showcasing it.

[harishsdev]

@mthrok Thanks for inputs

I have forked the repository and created Spectrogram branch with following commands rebased,please give me some inputs if i am wrong

git remote add upstream https://github.com/pytorch/audio.git
git fetch upstream

git checkout master

git rebase upstream/master

git push -f origin master

[harishsdev]

How to run flake in torchaudio for checking pep8 style coding, please suggest

[mthrok]

Hi @harishsdev

You can simple do flake8 torchaudio for flake8.

[mthrok]

1. This indentation is too much, there should be only 4 spaces instead of 8.
2. Please assign the transform into intermediate variable.
3. Please provide some arguments to Spectrogram constructor.
4. The input to Spectrogram should be called waveform.

[mthrok]

@harishsdev

Please build the doc by yourself or check the artifact of build_docs job.
The indentation is off.

https://239906-90321822-gh.circle-artifacts.com/0/docs/transforms.html#spectrogram

Also please, do assign the transform object to an intermediate variable to showcase the instantiation of Transform and application of the transform separately. Please do not write them in one line.

[mthrok]

1. Please fix the indentation. Build the doc and see that the rendering is correct.
2. Please assign the transform into intermediate variable.

[nateanl]

Hi @harishsdev. Could you assign the transform parameter and move the forward method to a separate line? You can see #1644 as an example.

transform = torchaudio.transforms.Spectrogram(n_fft = 512)
spectrogram = transform(waveform)

[harishsdev]

Sure I will update it

…

On Tue, Aug 3, 2021, 21:25 nateanl ***@***.***> wrote: ***@***.**** requested changes on this pull request. ------------------------------ In torchaudio/transforms.py <#1566 (comment)>: > @@ -67,6 +67,12 @@ class Spectrogram(torch.nn.Module): This argument is only effective when ``power=None``. It is ignored for cases where ``power`` is a number as in those cases, the returned tensor is power spectrogram, which is a real-valued tensor. + + Example + + >>> waveform, sample_rate = torchaudio.load('test.wav', normalization=True) + >>> transformed_spectrogram = torchaudio.transforms.Spectrogram(n_fft = 800)(waveform) Hi @harishsdev <https://github.com/harishsdev>. Could you assign the transform parameter and move the forward method to a separate line? You can see #1644 <#1644> as an example. transform = torchaudio.transforms.Spectrogram(n_fft = 512) spectrogram = transform(waveform) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1566 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALCDHPVBCAT63CGEZ3DK6QTT3AGOHANCNFSM46O6HLLQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email> .

[nateanl]

Same as #1586. We can merge it after addressing the nits.

[nateanl]

Suggested change

	>>> waveform, sample_rate = torchaudio.load('test.wav', normalization=True)
	>>> waveform, sample_rate = torchaudio.load('test.wav', normalize=True)

[nateanl]

Suggested change

	>>> transform = torchaudio.transforms.Spectrogram(n_fft = 800)
	>>> transform = torchaudio.transforms.Spectrogram(n_fft=800)

[nateanl]

Same for this one. Let's delete the line break.

Suggested change