Wav2vec2 Large Xls R 300m Br D2

🚀 wav2vec2-large-xls-r-300m-br-d2

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA - FOUNDATION/COMMON_VOICE_8_0 - BR dataset. It's designed for automatic speech recognition tasks, aiming to accurately transcribe speech in the Brazilian context.

✨ Features

• Fine - tuned Model: Based on the pre - trained facebook/wav2vec2-xls-r-300m, fine - tuned on the Brazilian Portuguese subset of the Common Voice 8.0 dataset.
• Evaluation Metrics: Achieves specific results on evaluation sets, including Loss and Wer metrics.

📚 Documentation

Model Information

Model Index

• Name: wav2vec2 - large - xls - r - 300m - br - d2
• Results:
  • Task 1: Speech Recognition on Common Voice 8 (BR)
    • Dataset: mozilla - foundation/common_voice_8_0 (BR)
    • Metrics:
      • Test WER: 0.49770598355954887
      • Test CER: 0.18090500890299605
  • Task 2: Automatic Speech Recognition on Robust Speech Event - Dev Data (BR)
    • Dataset: speech - recognition - community - v2/dev_data (BR)
    • Metrics:
      • Test WER: NA
      • Test CER: NA

Evaluation Results

This model achieves the following results on the evaluation set:

• Loss: 1.1257
• Wer: 0.4631

Evaluation Commands

1. To evaluate on mozilla - foundation/common_voice_8_0 with test split

python eval.py --model_id DrishtiSharma/wav2vec2-large-xls-r-300m-br-d2 --dataset mozilla-foundation/common_voice_8_0 --config br --split test --log_outputs

2. To evaluate on speech - recognition - community - v2/dev_data Breton language isn't available in speech - recognition - community - v2/dev_data

Training Hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.00034
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9, 0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 750
• num_epochs: 50
• mixed_precision_training: Native AMP

Training Results

Framework Versions

• Transformers 4.16.2
• Pytorch 1.10.0+cu111
• Datasets 1.18.3
• Tokenizers 0.11.0

📄 License

This model is released under the Apache - 2.0 license.