wav2vec2-arabic-colab

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

• Loss: 0.8713
• Wer: 49.0589
• Cer: 14.8792

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
6.1484	5.2685	200	3.0410	100.0	100.0
2.9396	10.5369	400	2.7504	100.0	100.0
0.9797	15.8054	600	0.7162	56.8517	17.6251
0.1969	21.0537	800	0.8082	50.9531	15.6761
0.0995	26.3221	1000	0.8713	49.0589	14.8792

Framework versions

• Transformers 4.50.0
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1