Skip to content

Fix HF transducer notebook generated model restore#15908

Open
KunalDhawan wants to merge 1 commit into
NVIDIA-NeMo:mainfrom
KunalDhawan:fix_asr_hf_notebook_restore
Open

Fix HF transducer notebook generated model restore#15908
KunalDhawan wants to merge 1 commit into
NVIDIA-NeMo:mainfrom
KunalDhawan:fix_asr_hf_notebook_restore

Conversation

@KunalDhawan

Copy link
Copy Markdown
Collaborator

What does this PR do ?

Fix Transducers_with_HF_Datasets.ipynb restore of the notebook-generated .nemo checkpoint.

Collection: ASR tutorials

Changelog

  • Remove the global TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1 override from the HF transducer notebook.
  • Clear the override before restoring telugu_asr_model.nemo so the generated state dict uses PyTorch's safe default load path.

Usage

No API changes.

Additional Information

  • Related to ASR tutorial checkpoint restore failure.

Signed-off-by: Kunal Dhawan <kunaldhawan97@gmail.com>
@KunalDhawan KunalDhawan requested a review from nithinraok July 13, 2026 17:23
@KunalDhawan KunalDhawan added the r3.0.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge. label Jul 13, 2026
@copy-pr-bot

copy-pr-bot Bot commented Jul 13, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@github-actions github-actions Bot added the ASR label Jul 13, 2026
@KunalDhawan

Copy link
Copy Markdown
Collaborator Author

/ok to test de1b4a4

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ASR r3.0.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant