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: 0.2522
• Wer: 23.1797

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: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.1114	0.0	1	2.3698	75.1864
0.3272	0.29	1000	0.4182	37.7505
0.251	0.58	2000	0.3408	30.9679
0.2207	0.88	3000	0.3059	28.3058
0.1779	1.17	4000	0.2890	26.7555
0.1691	1.46	5000	0.2742	25.2099
0.1622	1.75	6000	0.2645	24.6840
0.1397	2.04	7000	0.2587	23.8812
0.1394	2.34	8000	0.2562	23.6586
0.1361	2.63	9000	0.2536	23.4633
0.1356	2.92	10000	0.2522	23.1797

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0
• Datasets 2.11.0
• Tokenizers 0.13.3