omniASR-CTC-1B-v2
Wav2Vec2 CTC ASR model (v2) converted from the OmniLingual fairseq2 checkpoint omniASR_CTC_1B_v2.
This model outputs CTC logits over a SentencePiece vocabulary and can transcribe speech in multiple languages.
Code Base
The code base for the conversion can be found here. I was only able to convert the 300M and 1B models due to GPU limitations. Contributions are welcome.
Model details
| Property | Value |
|---|---|
| HF class | Wav2Vec2ForCTC |
| Encoder layers | 48 |
| Hidden size | 1280 |
| Attention heads | 16 |
| FFN intermediate | 5120 |
| Vocabulary size | 10288 |
| Source framework | fairseq2 |
| Source card | omniASR_CTC_1B_v2 |
| Parity verification | ✅ Verified |
Numerical parity against the original fairseq2 checkpoint has been confirmed: outputs match to within atol=1e-4 on a held-out audio sample.
Sample transcriptions on the held-out audio clip:
| Model | Transcript |
|---|---|
| fairseq2 (source) | concord returned to its place amidst the tents |
| HuggingFace (this repo) | concord returned to its place amidst the tents |
Usage
from transformers import Wav2Vec2ForCTC, AutoProcessor
import torch, torchaudio
processor = AutoProcessor.from_pretrained("aadel4/omniASR-CTC-1B-v2")
model = Wav2Vec2ForCTC.from_pretrained("aadel4/omniASR-CTC-1B-v2")
model.eval()
waveform, sr = torchaudio.load("audio.wav")
if sr != 16_000:
waveform = torchaudio.functional.resample(waveform, sr, 16_000)
inputs = processor(
waveform.squeeze().numpy(), sampling_rate=16_000, return_tensors="pt"
)
with torch.no_grad():
logits = model(**inputs).logits # (1, T, vocab)
pred_ids = torch.argmax(logits, dim=-1)
transcript = processor.decode(pred_ids[0])
print(transcript)