Conv1d subsampling

[burchim]

What does this PR do ?

Add conv1d subsampling for ASR

Collection: asr

Changelog

• Add 'striding_conv1d' and 'dw_striding_conv1d' subsampling options. Subsampling is performed using strided conv1d convolutions with kernel size 5 instead of doing reshape + frequency-temporal conv2d + reshape + linear projection.

Usage

• Simply use 'subsampling: striding_conv1d in conformer yaml config for example.

  encoder:
      _target_: nemo.collections.asr.modules.ConformerEncoder
      feat_in: ${model.video_front_end.dim_output}
      feat_out: -1 # you may set it if you need different output size other than the default d_model
      n_layers: 18
      d_model: 512
  
      # Sub-sampling params
      subsampling: striding_conv1d # vggnet, striding, stacking or stacking_norm, dw_striding, striding_conv1d_k5, dw_striding_conv1d
      subsampling_factor: 8 # must be power of 2 for striding and vggne
...

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

If you haven't finished some of the above items you can still open "Draft" PR.

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

• Related to # (issue)

[titu1994]

Looks ok to me, @nithinraok fyi