Removes use of torchaudio and moves transforms inside of NeMo

[blisc]

Important

The Update branch button must only be pressed in very rare occassions.
An outdated branch is never blocking the merge of a PR.
Please reach out to the automation team before pressing that button.

What does this PR do ?

Removes use of torchaudio.transforms and moves transforms inside of NeMo.
NOTE: we will use torchsquirm in nemo/collections/audio/metrics/squim.py and nemo/collections/tts/models/magpietts_preference_optimization.py

Collection: audio, asr, tts

Changelog

• Move frequently used torchaudio transform into NeMo

PR Type:

• New Feature
• Bugfix
• Documentation

[nithinraok]

Can we update or remove torchaudio references from following files as well:

• nemo/collections/tts/models/magpietts_preference_optimization.py
• docker/Dockerfile.speech
• scripts/installers/install_torchaudio_latest.sh
• docs/source/speechlm2/intro.rst
• nemo/collections/audio/metrics/squim.py

In following files they are being installed but not used. We can update them too.

• tutorials/00_NeMo_Primer.ipynb
• tutorials/01_NeMo_Models.ipynb
• tutorials/asr/Online_Offline_Microphone_VAD_Demo.ipynb
• tutorials/asr/Online_Offline_Speech_Commands_Demo.ipynb
• tutorials/asr/Streaming_Multitalker_ASR.ipynb
• tutorials/audio/speech_enhancement/BNR_Speech_enhancement_with_NeMo.ipynb
• tutorials/audio/speech_enhancement/Speech_Enhancement_with_NeMo.ipynb
• tutorials/speaker_tasks/ASR_with_SpeakerDiarization.ipynb
• tutorials/speaker_tasks/End_to_End_Diarization_Inference.ipynb
• tutorials/speaker_tasks/Speaker_Diarization_Inference.ipynb
• tutorials/speaker_tasks/Speaker_Identification_Verification.ipynb
• tutorials/speaker_tasks/Streaming_End_to_End_Diarization_Inference.ipynb

[nithinraok]

can we remove this option altogether?

[nithinraok]

I see from features.py FilterbankFeaturesTA doesn;t use torchaudio. I think we can remove this condition and default to FilterbankFeatures

[nithinraok]

why to remove these?

[blisc]

Causes a circular dependency if this stays in the init

[blisc]

Can we update or remove torchaudio references from following files as well:

• nemo/collections/tts/models/magpietts_preference_optimization.py
• docker/Dockerfile.speech
• scripts/installers/install_torchaudio_latest.sh
• docs/source/speechlm2/intro.rst
• nemo/collections/audio/metrics/squim.py

We do not have a replacement for SQIUM-MOS. I think we still need to keep the import check there. But we can remove it from the dockerfile and tutorials.

[pzelasko]

LGTM after addressing Nithin's concerns

[nithinraok]

@chtruong814 could you help review the docker file changes.

[nithinraok]

@blisc I think updates to docs/source/speechlm2/intro.rst are missing

[blisc]

@blisc I think updates to docs/source/speechlm2/intro.rst are missing

I don't know if there is an equivalent to torchaudio.load() so I'm deferring making that change. @PiotrDabkowski Do you have a recommendation on how to update that docs?

[nithinraok]

That failing CI test is okay to skip as its flaky. @chtruong814 FYI