Skip to content

Conversation

@artbataev
Copy link
Collaborator

@artbataev artbataev commented Jul 2, 2025

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An outdated branch is never blocking the merge of a PR.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 ?

  • clean up transducer decoding initialization
  • raise error if n-gram LM model is passed, but not supported by the strategy (previously silently ignored)

Collection: [ASR]

Changelog

  • Add specific line by line info of high level changes in this PR.

Usage

  • You can potentially add a usage example below
# Add a code snippet demonstrating how to use this 

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR.
To re-run CI remove and add the label again.
To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

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)

artbataev added 3 commits July 2, 2025 17:15
Signed-off-by: Vladimir Bataev <[email protected]>
Signed-off-by: Vladimir Bataev <[email protected]>
Signed-off-by: Vladimir Bataev <[email protected]>
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull Request Overview

Refactors transducer decoding initialization to use strongly-typed enums and a single match block for strategy/model logic.

  • Introduces TransducerModelType and TransducerDecodingStrategyType enums (via PrettyStrEnum).
  • Consolidates nested if-elif decoding branches into a match strategy, model_type structure.
  • Enhances error reporting for incompatible TDT and multi-blank configurations.
Comments suppressed due to low confidence (3)

nemo/collections/asr/parts/submodules/rnnt_decoding.py:488

  • The TSD case instantiates BeamRNNTInfer without passing search_type='tsd', so it will default to beam search rather than TSD. Please include the correct search_type argument or use a dedicated TSD inferer.
            case TransducerDecodingStrategyType.TSD, TransducerModelType.RNNT:

nemo/collections/asr/parts/submodules/rnnt_decoding.py:504

  • The ALSD branch uses BeamRNNTInfer but doesn't set search_type='alsd'. This likely runs standard beam search instead of ALSD. Add the search_type parameter or invoke an ALSD-specific infer class.
            case TransducerDecodingStrategyType.ALSD, TransducerModelType.RNNT:

nemo/collections/asr/parts/submodules/rnnt_decoding.py:593

  • The MAES_BATCH branch does not specify search_type='maes_batch' when creating BeamBatchedRNNTInfer, which can lead to incorrect default behavior. Consider adding the explicit search_type argument.
            case TransducerDecodingStrategyType.MAES_BATCH, TransducerModelType.RNNT:

Signed-off-by: Vladimir Bataev <[email protected]>
Copy link
Collaborator

@lilithgrigoryan lilithgrigoryan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Thank you!

@artbataev artbataev enabled auto-merge (squash) July 3, 2025 14:54
@github-actions github-actions bot removed the Run CICD label Jul 4, 2025
@github-actions
Copy link
Contributor

github-actions bot commented Jul 4, 2025

[🤖]: Hi @artbataev 👋,

We wanted to let you know that a CICD pipeline for this PR just finished successfully.

So it might be time to merge this PR or get some approvals.

//cc @chtruong814 @ko3n1g @pablo-garay @thomasdhc

@artbataev artbataev merged commit 07f96e2 into main Jul 4, 2025
247 checks passed
@artbataev artbataev deleted the cleanup_decoding_init branch July 4, 2025 12:47
AmirHussein96 pushed a commit to AmirHussein96/NeMo that referenced this pull request Jul 23, 2025
* Clean up strategy selection
* Raise error if LM not supported

---------

Signed-off-by: Vladimir Bataev <[email protected]>
Signed-off-by: Amir Hussein <[email protected]>
AmirHussein96 pushed a commit to AmirHussein96/NeMo that referenced this pull request Aug 5, 2025
* Clean up strategy selection
* Raise error if LM not supported

---------

Signed-off-by: Vladimir Bataev <[email protected]>
Signed-off-by: Amir Hussein <[email protected]>
AmirHussein96 pushed a commit to AmirHussein96/NeMo that referenced this pull request Aug 5, 2025
* Clean up strategy selection
* Raise error if LM not supported

---------

Signed-off-by: Vladimir Bataev <[email protected]>
Signed-off-by: Amir Hussein <[email protected]>
nasretdinovr pushed a commit to nasretdinovr/NeMo that referenced this pull request Aug 8, 2025
* Clean up strategy selection
* Raise error if LM not supported

---------

Signed-off-by: Vladimir Bataev <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants