Views
No views yet
nb-NO) ASR model, fine-tuned from
nvidia/nemotron-3.5-asr-streaming-0.6b on
LokaalHub/nb-NO-asr-cv.Community fine-tune, not an NVIDIA model. A derivative of NVIDIA's Nemotron 3.5 ASR. NVIDIA did not produce, endorse, or review this model. "Nemotron" is a trademark of NVIDIA, used here only to identify the base model.
nb-NO) is a supported locale of the base model, but its out-of-the-box accuracy on Common Voice is modest (~76.09% WER). A single full fine-tune on ~86.9h brings it to ~17.35% WER. Prompt slot used during fine-tuning: nb-NO (own slot).| Condition | Base | Fine-tuned | Rel. improvement |
|---|---|---|---|
WER (offline, full-context, normalized) on LokaalHub/nb-NO-asr-cv test | 76.09% | 17.35% | 77.2% |
Offline (full-context) WER via NeMotranscribe_speech.py. Cache-aware streaming WER (the condition NVIDIA headlines) was not measured for this release.
1import nemo.collections.asr as nemo_asr
2m = nemo_asr.models.ASRModel.restore_from("model.nemo") # from this repo
3m.transcribe(["audio.wav"]) # target_lang prompt: nb-NOinit_from_nemo_model), bf16, NoamAnnealing. Data:
LokaalHub/nb-NO-asr-cv (~86.9h train).
Built and trained by the asr-loop pipeline.