Wav2vec2 Large Xls R 300m Bg V1

🚀 wav2vec2-large-xls-r-300m-bg-v1

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the Bulgarian dataset of MOZILLA - FOUNDATION/COMMON_VOICE_8_0, designed for automatic speech recognition.

🚀 Quick Start

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA - FOUNDATION/COMMON_VOICE_8_0 - BG dataset. It achieves the following results on the evaluation set:

• Loss: 0.5197
• Wer: 0.4689

✨ Features

• Multilingual Adaptability: Based on the large - scale pre - trained model wav2vec2 - xls - r - 300m, it has good adaptability to different languages.
• High - precision Recognition: Demonstrates excellent performance in automatic speech recognition tasks on Bulgarian datasets.

📦 Installation

The README does not provide installation steps, so this section is skipped.

💻 Usage Examples

The README does not provide code examples, so this section is skipped.

📚 Documentation

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-bg-v1 --dataset mozilla-foundation/common_voice_8_0 --config bg --split test --log_outputs

2. To evaluate on speech - recognition - community - v2/dev_data

python eval.py --model_id DrishtiSharma/wav2vec2-large-xls-r-300m-bg-v1 --dataset speech-recognition-community-v2/dev_data --config bg --split validation --chunk_length_s 10 --stride_length_s 1

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 7e - 05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2000
• num_epochs: 50.0
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.17.0.dev0
• Pytorch 1.10.2+cu102
• Datasets 1.18.2.dev0
• Tokenizers 0.11.0

🔧 Technical Details

The model wav2vec2-large-xls-r-300m-bg-v1 is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the Bulgarian dataset of MOZILLA - FOUNDATION/COMMON_VOICE_8_0. Through fine - tuning, it can better adapt to the characteristics of Bulgarian speech, thereby improving the accuracy of speech recognition.

📄 License

This model is licensed under the Apache 2.0 license.

Information Table