Hanhpt23/whisper-small-Encode-engmed

openai/whisper-small

This model is a fine-tuned version of openai/whisper-small on the pphuc25/EngMed dataset. It achieves the following results on the evaluation set:

• Loss: 1.2775
• Wer: 25.8817

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: 0.0001
• train_batch_size: 8
• 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: 100
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8149	1.0	3491	0.8968	43.5550
0.5126	2.0	6982	0.8805	35.9730
0.3308	3.0	10473	0.9173	36.4008
0.2412	4.0	13964	0.9814	33.9519
0.1667	5.0	17455	1.0400	31.9450
0.1001	6.0	20946	1.0903	33.1316
0.079	7.0	24437	1.1191	33.2309
0.0751	8.0	27928	1.1580	29.3991
0.0515	9.0	31419	1.1723	29.8778
0.0395	10.0	34910	1.1876	29.4972
0.0313	11.0	38401	1.2305	28.8578
0.0129	12.0	41892	1.2344	27.5577
0.0207	13.0	45383	1.2527	29.9002
0.0169	14.0	48874	1.2498	27.3969
0.0055	15.0	52365	1.2604	27.4277
0.003	16.0	55856	1.2627	26.8923
0.0011	17.0	59347	1.2700	26.9135
0.0013	18.0	62838	1.2695	27.4785
0.0001	19.0	66329	1.2804	26.2836
0.0	20.0	69820	1.2775	25.8817

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1