Skip to content

Add experimental simulated chunked streaming decoding for transducers#15517

Draft
artbataev wants to merge 3 commits intomainfrom
vbataev/streaming_simulated
Draft

Add experimental simulated chunked streaming decoding for transducers#15517
artbataev wants to merge 3 commits intomainfrom
vbataev/streaming_simulated

Conversation

@artbataev
Copy link
Collaborator

@artbataev artbataev commented Mar 18, 2026

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An outdated branch is never blocking the merge of a PR.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

Add simulated decoding (False by default) to examples/asr/asr_chunked_inference/rnnt/speech_to_text_streaming_infer_rnnt.py script for faster experiments
and experiments with different decoding algorithms not yet implemented in streaming.

Approach:

  • encoder is evaluated on chunks (as in real streaming)
  • output is concatenated and decoded at one step (usually not used in real applications)

Expected to provide the same results if the decoding strategy supports streaming decoding without additional heuristics (e.g., pruning between steps).

Collection: [ASR]

Changelog

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

Usage

python examples/asr/asr_chunked_inference/rnnt/speech_to_text_streaming_infer_rnnt.py \
   pretrained_name="nvidia/parakeet-rnnt-1.1b" \
   dataset_manifest="test_other.jsonl" \
   batch_size=32 \
   output_filename="decoded.jsonl" \
   compute_dtype=bfloat16 \
   use_per_stream_biasing=false \
   calculate_rtfx=true \
   left_context_secs=10 \
   chunk_secs=0.16 \
   right_context_secs=0.8 \
   simulated=true \
   decoding.strategy=malsd_batch \
   decoding.beam.beam_size=32

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The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.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

@github-actions github-actions bot added the ASR label Mar 18, 2026
@artbataev artbataev changed the title ASR transducer decoding: add simulated chunked streaming decoding Add experimental simulated chunked streaming decoding for transducers Mar 18, 2026
Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
@artbataev artbataev force-pushed the vbataev/streaming_simulated branch from 7939dd8 to 5c809e3 Compare March 18, 2026 17:26
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.

1 participant