mayabedge/whisper-ft

Whisper Fine-tuned - NNCES

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

• Loss: 0.1135
• Wer: 8.0963

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: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2697	0.1	10	0.8252	40.9920
0.6597	0.2	20	0.5482	25.2371
0.4656	0.3	30	0.3488	20.0584
0.2774	0.4	40	0.2164	21.5901
0.1746	0.5	50	0.1770	19.0372
0.1826	0.6	60	0.1540	15.3902
0.1228	0.7	70	0.1364	11.4515
0.1271	0.8	80	0.1246	8.6798
0.2388	0.9	90	0.1165	8.0233
0.2584	1.0	100	0.1135	8.0963

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1