whisper-base

This model is a fine-tuned version of openai/whisper-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1952
• Wer: 10.4439

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: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 60000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2139	0.0833	5000	0.1884	16.6399
0.1146	0.1667	10000	0.1447	13.0148
0.0686	0.25	15000	0.1384	11.3586
0.0427	0.3333	20000	0.1471	11.4970
0.0274	0.4167	25000	0.1585	10.8926
0.0195	0.5	30000	0.1702	11.3447
0.0155	0.5833	35000	0.1773	10.6100
0.0126	1.0062	40000	0.1863	11.4255
0.0099	1.0895	45000	0.1929	10.6665
0.01	1.1729	50000	0.1933	10.6665
0.0085	1.2562	55000	0.1953	10.5224
0.0085	1.3395	60000	0.1952	10.4439

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.20.3