Marialab/whisper-small-dr-ar

Whisper small darija translate

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

• Loss: 0.0001
• Bleu: 0.6434

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Use 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: 2000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
0.7235	133.3333	2000	0.5970	0.05
0.0257	266.6667	4000	0.0500	0.6200
0.0003	400.0	6000	0.0002	0.7200
0.0001	533.3333	8000	0.0001	0.6567
0.0001	666.6667	10000	0.0001	0.6434

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.0+cu121
• Datasets 2.19.2
• Tokenizers 0.20.3