whisper-medium-toi

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: 1.0796
• Wer: 35.2601

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.6522	0.24	500	2.0369	75.7050
0.9481	0.48	1000	1.3940	48.5549
0.6936	0.72	1500	1.2731	44.5262
0.6486	0.96	2000	1.1436	40.5500
0.6288	1.2	2500	1.1495	38.6057
0.5257	1.44	3000	1.1033	37.1519
0.4218	1.68	3500	1.0615	36.3461
0.4935	1.92	4000	1.0796	35.2601

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2