Hanhpt23/whisper-tiny-engmed-v2

openai/whisper-tiny

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

• Loss: 1.6241
• Wer: 29.3019

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
0.8114	1.0	2266	0.8947	35.8609
0.5582	2.0	4532	0.9096	44.9746
0.3646	3.0	6798	0.9616	33.8310
0.2497	4.0	9064	1.0257	35.3593
0.1953	5.0	11330	1.1146	33.6587
0.1151	6.0	13596	1.2047	35.6933
0.083	7.0	15862	1.2631	32.3759
0.068	8.0	18128	1.3303	33.5213
0.0389	9.0	20394	1.3732	31.8373
0.0272	10.0	22660	1.4187	32.7587
0.0244	11.0	24926	1.4516	32.2493
0.0173	12.0	27192	1.4985	32.1704
0.0109	13.0	29458	1.5192	31.6727
0.0045	14.0	31724	1.5456	30.2837
0.0043	15.0	33990	1.5739	30.1902
0.0017	16.0	36256	1.5916	30.6646
0.0006	17.0	38522	1.6012	29.8425
0.0002	18.0	40788	1.6050	29.3857
0.0013	19.0	43054	1.6176	29.7792
0.0001	20.0	45320	1.6241	29.3019

Framework versions

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