whisper-medium-order

This model is a fine-tuned version of nandovallec/whisper-tiny-bg-l on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0173
• Wer: 0.0

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: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 64
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.624	3.0	5	1.4084	111.7647
0.5102	6.0	10	0.5733	43.1373
0.2173	8.78	15	0.3225	31.3725
0.0178	11.29	20	0.2035	27.4510
0.0138	14.0	25	0.1143	21.5686
0.036	17.0	30	0.0603	3.9216
0.0296	20.0	35	0.0366	1.9608
0.0074	22.59	40	0.0250	1.9608
0.0014	25.1	45	0.0191	0.0
0.0051	28.0	50	0.0173	0.0

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0