Arbi-Houssem/TunLangModel1.6

Whisper Tunisien

This model is a fine-tuned version of openai/whisper-small on the Tunisian_dataset_STT-TTS1 dataset. It achieves the following results on the evaluation set:

• Loss: 2.9262
• Wer: 99.4163

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: 1e-06
• train_batch_size: 8
• 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: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.565	10.3093	1000	2.7181	99.5460
0.2875	20.6186	2000	2.7486	106.3554
0.1701	30.9278	3000	2.8744	103.2425
0.1375	41.2371	4000	2.9262	99.4163

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1