[ASR] k2 rnnt

[GNroy]

What does this PR do ?

Adds graph-based RNNT support.

Collection: ASR

Changelog

• k2 losses and decoders refactored
• graph-based RNNT added
• RNNT greedy and forced algnment added
• CTC and RNNT tests added

Usage

• RNNT forced alignment: python ./experimental/k2/align_speech_parallel.py model=stt_en_conformer_transducer_large predict_ds.manifest_filepath=./data/Librispeech/dev_clean.json predict_ds.batch_size=8 output_path=./ali_rnnt model_stride=4 aligner_args.decode_batch_size=2 aligner_args.rnnt_cfg.predictor_window_size=40 aligner_args.rnnt_cfg.predictor_step_size=4

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:

• [x] New Feature
• [ ] Bugfix
• [ ] Documentation

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

Who can review?

@titu1994 Might be interesting for @erastorgueva-nv @yzhang123 @artbataev Contributor guidelines contains specific people who can review PRs to various areas.

[lgtm-com[bot]]

This pull request introduces 5 alerts when merging 868032b878f1a59a4139ab022949927e90edf8d5 into 4aba4b208cf955660759e0c0cd07057a6049ffc8 - view on LGTM.com

new alerts:

• 2 for First parameter of a method is not named 'self'
• 1 for Wrong name for an argument in a call
• 1 for Unused import
• 1 for Wrong number of arguments in a call

[lgtm-com[bot]]

This pull request introduces 1 alert when merging 4329fcda62e21824f0696af2d5537acd2293c3f9 into 4aba4b208cf955660759e0c0cd07057a6049ffc8 - view on LGTM.com

new alerts:

• 1 for `__init__` method calls overridden method

[lgtm-com[bot]]

This pull request introduces 1 alert when merging d9e0f64671fcc484ef077f6f07e247f1fd0d680b into ddd87197e94ca23ae54e641dc7784e64c00a43d6 - view on LGTM.com

new alerts:

• 1 for `__init__` method calls overridden method

[lgtm-com[bot]]

This pull request introduces 4 alerts when merging 094fd307d903f4db57181d58b695d9d66fc4434e into 75c2d8258106aec4e7c136df35c564f3c030575f - view on LGTM.com

new alerts:

• 2 for Unused import
• 1 for Unused local variable
• 1 for `__init__` method calls overridden method

[lgtm-com[bot]]

This pull request introduces 1 alert when merging bf53fc91f5a91f976cb648edd8024d446133e123 into 28b97af8bd02a57075c56de483f16b6f0a8b6b11 - view on LGTM.com

new alerts:

• 1 for `__init__` method calls overridden method

[lgtm-com[bot]]

This pull request introduces 1 alert when merging 8143205ed07a3c4e7cf7c4c430b973b246a11e31 into 28b97af8bd02a57075c56de483f16b6f0a8b6b11 - view on LGTM.com

new alerts:

• 1 for `__init__` method calls overridden method

[lgtm-com[bot]]

This pull request introduces 1 alert when merging f2c13ae5e30ee9c20ddd404aa492bd6ce8db127f into 28b97af8bd02a57075c56de483f16b6f0a8b6b11 - view on LGTM.com

new alerts:

• 1 for `__init__` method calls overridden method

[lgtm-com[bot]]

This pull request introduces 1 alert when merging 73448b966ecde65a879dd1ec381ba7218ecc7aec into 28b97af8bd02a57075c56de483f16b6f0a8b6b11 - view on LGTM.com

new alerts:

• 1 for `__init__` method calls overridden method

[artbataev]

Regarding repeated snippet

# use k2 import guard
# fmt: off
from nemo.core.utils.k2_utils import k2_import_guard # isort:skip
k2_import_guard()
import k2 # isort:skip
# fmt: on

Maybe it would be better to place it into __init__.py, and remove from other files in the module (just to avoid duplicated fragments)

[github-actions[bot]]

This PR is stale because it has been open for 30 days with no activity. Remove stale label or comment or update or this will be closed in 7 days.

[github-actions[bot]]

This PR was closed because it has been inactive for 7 days since being marked as stale.

[github-actions[bot]]

This PR is stale because it has been open for 14 days with no activity. Remove stale label or comment or update or this will be closed in 7 days.

[github-actions[bot]]

This PR was closed because it has been inactive for 7 days since being marked as stale.