Make possible to control tqdm progress bar in ASR models

[SN4KEBYTE]

What does this PR do ?

Now it's possible to control display of progress bar when calling transcribe method of ASR models.

Collection: ASR

Changelog

• transcribe method of ASR models now have verbose param, which allows to control display of tqdm progress bar

Usage

import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.EncDecRNNTBPEModel.from_pretrained("nvidia/stt_en_conformer_transducer_xlarge")

# Will display progress bar by default
asr_model.transcribe(['2086-149220-0033.wav'])

# Won't display progress bar when verbose is set to False
asr_model.transcribe(['2086-149220-0033.wav'], verbose=False)

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
  • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• New Feature
• Bugfix
• Documentation

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

None

[titu1994]

Sorry but this is by design. We don't want to disable tqdm with special args because that would need to update every single Model with those args in its transcribe function.

There are other ways to disable logging at least locally - #1376 (comment)

But tqdm wont be updated for now.

[titu1994]

Ah to note - if you really need this, this is a viable alternative - https://stackoverflow.com/a/67238486

[titu1994]

On second thought, let me discuss with the team. Might be a nice arg to add. We'll get back to you in some time.

[titu1994]

After discussion, seems the team is in favor of adding this.
Thanks for the PR !

[SN4KEBYTE]

Thank you! I was happy to help.