Wav2vec2 Large Xls R 300m Breton Cv8

🚀 XLS-R-300M - Breton

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA - FOUNDATION/COMMON_VOICE_8_0 - BR dataset. It is designed for automatic speech recognition tasks and has achieved certain results in relevant evaluations.

📚 Documentation

Model Information

Evaluation Results

This model achieves the following results on the evaluation set:

• Loss: NA
• Wer: NA

In the model index, for the Automatic Speech Recognition task on the Common Voice 8 dataset (type: mozilla - foundation/common_voice_8_0, args: br), the metrics are as follows:

• Test WER: 54.855
• Test CER: 17.865

Framework Versions

• Transformers 4.16.0.dev0
• Pytorch 1.10.0+cu102
• Datasets 1.17.1.dev0
• Tokenizers 0.10.3

💻 Usage Examples

Evaluation Commands

Evaluate on mozilla - foundation/common_voice_8_0 with split test

python eval.py --model_id infinitejoy/wav2vec2-large-xls-r-300m-breton-cv8 --dataset mozilla-foundation/common_voice_8_0 --config br --split test

Evaluate on speech - recognition - community - v2/dev_data

python eval.py --model_id infinitejoy/wav2vec2-large-xls-r-300m-breton-cv8 --dataset speech-recognition-community-v2/dev_data --config br --split validation --chunk_length_s 5.0 --stride_length_s 1.0

Inference With LM

import torch
from datasets import load_dataset
from transformers import AutoModelForCTC, AutoProcessor
import torchaudio.functional as F


model_id = "infinitejoy/wav2vec2-large-xls-r-300m-breton-cv8"

sample_iter = iter(load_dataset("mozilla-foundation/common_voice_8_0", "br", split="test", streaming=True, use_auth_token=True))

sample = next(sample_iter)
resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), 48_000, 16_000).numpy()

model = AutoModelForCTC.from_pretrained(model_id)
processor = AutoProcessor.from_pretrained(model_id)

input_values = processor(resampled_audio, return_tensors="pt").input_values

with torch.no_grad():
    logits = model(input_values).logits

transcription = processor.batch_decode(logits.numpy()).text

Eval results on Common Voice 7 "test" (WER):

NA

📄 License

This model is released under the Apache - 2.0 license.