Wav2vec2 Large Xls R 300m Sat A3

🚀 wav2vec2-large-xls-r-300m-sat-a3

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA - FOUNDATION/COMMON_VOICE_8_0 - SAT dataset. It is mainly used for automatic speech recognition, aiming to provide high - quality speech - to - text conversion and perform well in related tasks.

✨ Features

• Fine - tuned Model: Based on the pre - trained model [facebook/wav2vec2-xls-r-300m], it is fine - tuned on the SAT dataset of MOZILLA - FOUNDATION/COMMON_VOICE_8_0 to adapt to specific speech recognition tasks.
• Multi - metric Evaluation: Evaluated using multiple metrics such as WER (Word Error Rate) and CER (Character Error Rate) to comprehensively measure the performance of the model.

📚 Documentation

Model Information

Evaluation Results

The model achieves the following results on the evaluation set:

• Loss: 0.8961
• Wer: 0.3976

The detailed evaluation results on different datasets are as follows:

Evaluation Commands

1. Evaluate on mozilla - foundation/common_voice_8_0 with test split

python eval.py --model_id DrishtiSharma/wav2vec2-large-xls-r-300m-sat-a3 --dataset mozilla-foundation/common_voice_8_0 --config sat --split test --log_outputs

2. Evaluate on speech - recognition - community - v2/dev_data Note: Santali (Ol Chiki) language not found in speech - recognition - community - v2/dev_data

Training Hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0004
• 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: 200
• num_epochs: 200
• 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