Hanhpt23/whisper-base-engmed-v2

openai/whisper-base

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

• Loss: 1.4040
• Wer: 24.2615

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
0.7995	1.0	2268	0.8023	42.7597
0.4972	2.0	4536	0.8141	31.7847
0.3198	3.0	6804	0.8778	34.2508
0.2148	4.0	9072	0.9436	32.1119
0.1419	5.0	11340	1.0230	29.0506
0.0918	6.0	13608	1.0874	27.6792
0.0747	7.0	15876	1.1259	28.6260
0.0515	8.0	18144	1.1792	28.2198
0.0271	9.0	20412	1.2107	26.5981
0.0248	10.0	22680	1.2557	25.6329
0.0234	11.0	24948	1.2852	25.0377
0.0123	12.0	27216	1.2924	25.9923
0.0078	13.0	29484	1.3419	26.1364
0.0066	14.0	31752	1.3504	25.6874
0.003	15.0	34020	1.3602	23.9576
0.0028	16.0	36288	1.3710	24.7884
0.0045	17.0	38556	1.3845	24.4455
0.0003	18.0	40824	1.3914	25.2929
0.0002	19.0	43092	1.3993	24.3930
0.0001	20.0	45360	1.4040	24.2615

Framework versions

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