Moroccan-Darija-STT-small-v1.6.6

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

• Loss: 0.4661
• Wer: 109.2620
• Cer: 69.5674

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: 1.25e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• 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: 10
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.2636	0.6098	25	0.4860	125.0753	82.7965
0.8777	1.2195	50	0.4632	98.9793	63.2755
0.8238	1.8293	75	0.4932	112.4749	73.8932
0.8339	2.4390	100	0.4695	112.9936	75.7698
0.8101	3.0488	125	0.4685	115.8635	72.3950
0.7691	3.6585	150	0.4753	131.5177	88.1459
0.7458	4.2683	175	0.4738	105.2962	67.1807
0.7414	4.8780	200	0.4578	106.0576	64.3092
0.6999	5.4878	225	0.4661	109.2620	69.5674

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0