Hanhpt23/whisper-medium-engmed-v2

openai/whisper-medium

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

• Loss: 1.3051
• Wer: 19.0281

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
0.8152	1.0	2268	0.7891	34.2392
0.4587	2.0	4536	0.7773	24.6287
0.3061	3.0	6804	0.8282	23.0274
0.2015	4.0	9072	0.8892	22.5628
0.1347	5.0	11340	0.9453	19.5775
0.0886	6.0	13608	1.0157	20.1560
0.0726	7.0	15876	1.0550	19.8638
0.0563	8.0	18144	1.0648	19.7586
0.0421	9.0	20412	1.0888	21.6560
0.0222	10.0	22680	1.1299	20.5330
0.0313	11.0	24948	1.1475	19.1314
0.0168	12.0	27216	1.1656	19.4918
0.0212	13.0	29484	1.1961	19.7869
0.0093	14.0	31752	1.2353	19.5599
0.0036	15.0	34020	1.2368	20.1463
0.0026	16.0	36288	1.2582	19.0077
0.0034	17.0	38556	1.2643	19.4324
0.0012	18.0	40824	1.2775	19.5064
0.0002	19.0	43092	1.2977	19.3009
0.0008	20.0	45360	1.3051	19.0281

Framework versions

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