wav2vec2-base-10k-voxpopuli-ft-it-onnx

wav2vec2-base-10k-voxpopuli-ft-it-onnx (ONNX)

⚠️ Non-commercial use only. This model is licensed CC BY-NC 4.0, inherited from the original Meta AI checkpoint. Unlike most other onnx-asr conversions in this collection (Apache-2.0 / CC0 / CC-BY), this model may not be used for commercial purposes.

ONNX export of facebook/wav2vec2-base-10k-voxpopuli-ft-it, a Italian wav2vec2 CTC ASR model from Meta AI (FAIR) — VoxPopuli project, fine-tuned on the VoxPopuli European Parliament speech corpus — for use with onnx-asr (wav2vec2-ctc model type).

Per-utterance zero-mean/unit-variance normalization is baked into the ONNX graph, masked by input_lengths for correct behavior with padded/batched input, so the model works with onnx-asr's plain identity preprocessor (raw 16kHz waveform in).

Usage

import onnx_asr

model = onnx_asr.load_model("OpenVoiceOS/wav2vec2-base-10k-voxpopuli-ft-it-onnx")
print(model.recognize("test.wav"))

Files

• model.onnx / model.onnx.data — fp32 ONNX graph (inputs: input_values (batch, samples) float32, input_lengths (batch,) int64; output: logprobs (batch, frames, vocab) float32 log-softmax).
• vocab.txt — CTC vocabulary in onnx-asr's token id format (word-delimiter -> ▁, pad token -> <blk>).
• config.json — {"model_type": "wav2vec2-ctc", "subsampling_factor": 320}.

No int8 quantized variant is included yet -- onnxruntime.quantization does not currently support the torch.onnx dynamo-exported graph for this architecture.

License

CC BY-NC 4.0 — non-commercial use only. Inherited from the base model (facebook/wav2vec2-base-10k-voxpopuli-ft-it, Meta AI / FAIR). See the license text for full terms.