Hanhpt23/whisper-small-engmed-v2

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: 1.2969
• Wer: 19.8132

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
0.7724	1.0	2268	0.7573	43.9665
0.439	2.0	4536	0.7693	43.3383
0.2876	3.0	6804	0.8212	32.4148
0.1882	4.0	9072	0.8900	25.3134
0.128	5.0	11340	0.9573	26.0614
0.0812	6.0	13608	1.0126	21.0453
0.0635	7.0	15876	1.0505	22.0544
0.0471	8.0	18144	1.0961	22.6690
0.032	9.0	20412	1.1188	21.7300
0.0238	10.0	22680	1.1409	22.2326
0.0228	11.0	24948	1.1615	21.0589
0.0165	12.0	27216	1.2063	21.2099
0.0073	13.0	29484	1.2233	20.5515
0.0088	14.0	31752	1.2199	20.3353
0.0037	15.0	34020	1.2605	20.1151
0.0021	16.0	36288	1.2565	19.8531
0.0044	17.0	38556	1.2666	20.6187
0.0006	18.0	40824	1.2941	19.7616
0.0001	19.0	43092	1.2905	19.9700
0.0001	20.0	45360	1.2969	19.8132

Framework versions

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