whisper-v3-30s-segments

This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.0606
• Wer: 37.0618

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6589	1.0	9	0.6000	35.3088
0.35	2.0	18	0.5675	39.0651
0.2118	3.0	27	0.6486	36.1436
0.1167	4.0	36	0.6838	39.7329
0.0621	5.0	45	0.8158	37.1452
0.031	6.0	54	0.9053	36.6444
0.0154	7.0	63	0.9873	38.3973
0.0085	8.0	72	1.0745	38.2304
0.0059	8.9697	80	1.0606	37.0618

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0