fix(nemo/qnn/parakeet-ctc): wrapper.py raises NameError on undefined feat_dim

[Anai-Guo]

scripts/nemo/qnn/parakeet-ctc/wrapper.py uses feat_dim without ever binding it:

    m = ModelWrapper(asr_model)
    m.eval()

    print("feat_dim", feat_dim)                                    # NameError
    x = torch.rand(1, feat_dim, args.max_len, dtype=torch.float32)

$ pyflakes scripts/nemo/qnn/parakeet-ctc/wrapper.py
scripts/nemo/qnn/parakeet-ctc/wrapper.py:67:23: undefined name 'feat_dim'
scripts/nemo/qnn/parakeet-ctc/wrapper.py:68:23: undefined name 'feat_dim'

Those are the only two undefined names in the file, and they sit on the straight-line path of
main(), so python wrapper.py --max-len ... --model-id ... always dies right after downloading
and printing the model — before it exports anything.

The sibling script one directory over,
scripts/nemo/qnn/nemotron-3.5-asr-streaming-0.6b/wrapper.py:370, has the line this one is
missing:

    feat_dim = asr_model.cfg.preprocessor.features
    ...
    x = torch.rand(1, feat_dim, window_size, dtype=torch.float32)

(scripts/nemo/nemotron-3.5-asr-streaming-0.6b/export_onnx.py:186 does the same.)

Reproduction

NeMo is not needed to show it — main is lifted out of the file with ast and run against a
stub whose cfg behaves like the OmegaConf config NeMo returns
(preprocessor.features = 128), with torch.onnx.export captured:

### python wrapper.py --max-len 200 --model-id nvidia/parakeet-ctc-0.6b
  [before] NameError: name 'feat_dim' is not defined
  [after]  feat_dim 128
           exported input shape (1, 128, 200)

(1, 128, 200) is (N, C, T) as documented in ModelWrapper.forward.

Fix

Add the missing feat_dim = asr_model.cfg.preprocessor.features, matching the sibling wrapper.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Improved ONNX export compatibility by matching the generated input tensor to the model’s configured feature dimensions.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: b494f902-3848-4bb7-8895-206d02d42f31

📥 Commits

Reviewing files that changed from the base of the PR and between f64f90b and bba9272.

📒 Files selected for processing (1)

• scripts/nemo/qnn/parakeet-ctc/wrapper.py

Included review availability: Your plan provides up to 10 included reviews per hour; 9 remain after this review.

---

📝 Walkthrough

Walkthrough

The ONNX export wrapper now reads the model’s configured preprocessor feature dimension before constructing the random input tensor.

Changes

ONNX export input configuration

Layer / File(s)	Summary
Read configured feature dimension scripts/nemo/qnn/parakeet-ctc/wrapper.py	main() retrieves asr_model.cfg.preprocessor.features for input-tensor construction.

Priority: ⬇️ Low

Estimated code review effort: 1 (Trivial) | ~2 minutes

Merge Risk: ⚪ Minimal · up to bba92

The export wrapper now obtains the configured feature dimension before creating its input tensor, preventing the prior NameError and allowing ONNX export to proceed. No current merge-blocking risk is identified.

Suggested reviewers: csukuangfj

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 1 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the affected wrapper and the primary fix: defining the undefined feat_dim that caused a NameError.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[csukuangfj]

Thank you for your contribution!