whisper-tiny-en

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8729
• Wer Ortho: 0.3344
• Wer: 0.3377

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: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0006	17.8571	500	0.6617	0.3251	0.3264
0.0002	35.7143	1000	0.7217	0.3257	0.3270
0.0001	53.5714	1500	0.7577	0.3226	0.3247
0.0001	71.4286	2000	0.7870	0.3337	0.3347
0.0	89.2857	2500	0.8109	0.3325	0.3341
0.0	107.1429	3000	0.8329	0.3356	0.3377
0.0	125.0	3500	0.8529	0.3344	0.3371
0.0	142.8571	4000	0.8729	0.3344	0.3377

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1