Arbi-Houssem/TunLangModel

Whisper Tunisien

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

• Loss: 6.9324
• Wer: 100.4052

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-05
• train_batch_size: 16
• eval_batch_size: 8
• 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: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.5017	100.0	300	6.1135	163.5332
0.0224	200.0	600	6.5103	105.8347
0.0101	300.0	900	6.8122	105.8347
0.0095	400.0	1200	6.8766	100.0
0.0093	500.0	1500	6.9174	100.0
0.0091	600.0	1800	6.9324	100.4052

Framework versions

• Transformers 4.41.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1