Naji888/whisper-tiny-en_v6

Whisper tiny En v6 Naji

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7726
• Wer Ortho: 32.1991
• Wer: 23.5413

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: 3e-07
• train_batch_size: 16
• eval_batch_size: 16
• 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: constant_with_warmup
• lr_scheduler_warmup_steps: 100
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
3.6998	0.0988	60	3.8951	33.0775	25.0106
3.18	0.1977	120	3.0495	33.0341	24.8557
2.1622	0.2965	180	2.3747	32.9327	24.9026
1.607	0.3954	240	1.7963	32.0156	23.6915
0.9201	0.4942	300	1.2851	32.2184	23.5178
0.6372	0.5931	360	1.0129	33.5505	24.9824
0.4641	0.6919	420	0.8403	32.5080	23.8276
0.433	0.7908	480	0.7726	32.1991	23.5413

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 2.14.6
• Tokenizers 0.21.1