whisper-medium-mn-4

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6015
• Wer: 33.0293
• Cer: 10.9236

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: 32
• eval_batch_size: 16
• 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: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0362	4.26	1000	0.4204	40.2720	13.8389
0.0087	8.51	2000	0.4712	37.4918	12.9175
0.0044	12.77	3000	0.4893	36.3393	12.4727
0.0033	17.02	4000	0.5159	35.8423	12.2933
0.0017	21.28	5000	0.5183	35.2797	12.1104
0.0016	25.53	6000	0.5422	35.4326	11.7454
0.0011	29.79	7000	0.5361	34.5314	11.5196
0.0004	34.04	8000	0.5406	34.0998	11.3650
0.0006	38.3	9000	0.5540	33.8650	11.2912
0.0002	42.55	10000	0.5748	34.0889	11.5333
0.0003	46.81	11000	0.5771	34.5641	11.4895
0.0	51.06	12000	0.5809	33.4335	11.2070
0.0	55.32	13000	0.5941	33.2095	11.0009
0.0	59.57	14000	0.6015	33.0293	10.9236
0.0	63.83	15000	0.6045	33.0347	10.9125

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2