wav2vec2-arabic-colab-f

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the audiofolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.8010
• Wer: 32.0897
• Cer: 9.9979

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• 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: 300
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.0756	13.1611	500	0.5087	41.6867	12.5220
0.3284	26.3221	1000	0.6221	36.1204	11.0568
0.1148	39.4832	1500	0.6592	34.9208	10.7164
0.0781	52.6443	2000	0.7070	33.2774	10.2609
0.0589	65.8054	2500	0.7958	32.5336	10.0942
0.0502	78.9664	3000	0.8010	32.0897	9.9979

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1