Skip to content

Rename NemotronCacheConfig to NemotronConfig and add blank penalty to the decoder - #2042

Merged
nenad1002 merged 4 commits into
mainfrom
nebanfic/nemotron-blank-penalty
Apr 16, 2026
Merged

Rename NemotronCacheConfig to NemotronConfig and add blank penalty to the decoder#2042
nenad1002 merged 4 commits into
mainfrom
nebanfic/nemotron-blank-penalty

Conversation

@nenad1002

Copy link
Copy Markdown
Contributor

No description provided.

Copilot AI review requested due to automatic review settings March 22, 2026 18:42

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Pull request overview

This PR updates the Nemotron Speech (RNNT streaming ASR) model integration by renaming the model config struct and wiring a new blank_penalty parameter from genai_config.json into the RNNT decoding step.

Changes:

  • Renamed NemotronCacheConfig to NemotronConfig across the Nemotron speech and streaming processor code.
  • Added blank_penalty to Config::Model, parsed it from JSON, and propagated it into NemotronConfig.
  • Applied blank_penalty during joiner-logits argmax to discourage (or encourage) blank emissions.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file
File Description
src/models/nemotron_streaming_processor.h Switched getters/member from cache_config_ to nemotron_config_.
src/models/nemotron_streaming_processor.cpp Uses nemotron_config_ throughout for streaming mel + cache behavior.
src/models/nemotron_speech.h Renamed config struct to NemotronConfig and added blank_penalty.
src/models/nemotron_speech.cpp Propagates blank_penalty from config and applies it during decoding argmax.
src/config.h Added blank_penalty field under streaming ASR / RNNT parameters.
src/config.cpp Added JSON parsing for model.blank_penalty.

Comment thread src/models/nemotron_speech.cpp
Comment thread src/models/nemotron_speech.cpp Outdated
Comment thread src/models/nemotron_speech.cpp Outdated
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants