Dev372/Medical_base_en_1_1v_batch_2x

English Whisper Model

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

• Loss: 0.1386
• Wer: 6.5809

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: 1e-05
• train_batch_size: 36
• 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: 500
• training_steps: 1500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1731	0.5650	100	0.9844	10.2812
0.6483	1.1299	200	0.6288	9.3047
0.3802	1.6949	300	0.3554	7.8938
0.1437	2.2599	400	0.1702	7.1230
0.1136	2.8249	500	0.1415	6.5841
0.0752	3.3898	600	0.1336	6.0616
0.0713	3.9548	700	0.1257	6.1236
0.0373	4.5198	800	0.1279	5.8526
0.0311	5.0847	900	0.1283	5.8003
0.03	5.6497	1000	0.1303	6.1171
0.0166	6.2147	1100	0.1314	6.0845
0.0241	6.7797	1200	0.1339	6.3588
0.0164	7.3446	1300	0.1368	6.3555
0.0178	7.9096	1400	0.1380	6.4764
0.0099	8.4746	1500	0.1386	6.5809

Framework versions

• Transformers 4.44.0
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1