results

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

• Loss: 0.0893
• Wer: 0.0991

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: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 500
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9767	1.0	7	0.4806	0.2243
0.3635	2.0	14	0.2216	0.1310
0.1596	3.0	21	0.1460	0.1016
0.076	4.0	28	0.1198	0.0915
0.0405	5.0	35	0.1142	0.0906
0.0262	6.0	42	0.1034	0.0901
0.0151	7.0	49	0.0965	0.1019
0.0083	8.0	56	0.0924	0.0963
0.0043	9.0	63	0.0899	0.0991
0.0023	10.0	70	0.0902	0.1019
0.0014	11.0	77	0.0924	0.1046
0.0012	12.0	84	0.0923	0.1040
0.0009	13.0	91	0.0915	0.1040
0.0008	14.0	98	0.0907	0.1040
0.0007	15.0	105	0.0904	0.0957
0.0007	16.0	112	0.0888	0.0957
0.0006	17.0	119	0.0900	0.0957
0.0006	18.0	126	0.0895	0.0957
0.0005	19.0	133	0.0894	0.0991
0.0006	20.0	140	0.0893	0.0991

Framework versions

• Transformers 4.50.2
• Pytorch 2.4.0+cu121
• Datasets 3.4.1
• Tokenizers 0.21.0