arun100/whisper-base-fr-1

Whisper Base French

This model is a fine-tuned version of openai/whisper-base on the mozilla-foundation/common_voice_16_0 fr dataset. It achieves the following results on the evaluation set:

• Loss: 0.5654
• Wer: 27.6510

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 7000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.739	0.07	500	0.7506	35.0088
0.6131	1.07	1000	0.6595	31.4298
0.5311	2.07	1500	0.6301	30.6233
0.551	3.07	2000	0.6141	29.7819
0.4588	4.07	2500	0.6003	29.2527
0.4163	5.07	3000	0.5936	29.0292
0.4553	6.07	3500	0.5838	28.4799
0.4395	7.07	4000	0.5783	28.2488
0.4233	8.07	4500	0.5747	28.0779
0.4204	9.07	5000	0.5712	28.1122
0.4378	10.06	5500	0.5695	28.0578
0.4337	11.06	6000	0.5673	27.7817
0.4277	12.06	6500	0.5658	27.6634
0.419	13.06	7000	0.5654	27.6510

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0