meiiny00/whisper-tinyyy-better-total

Whisper ti qt

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

• Loss: 0.3074
• Wer: 37.6797

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4805	1.3353	1000	0.6147	65.1001
0.9661	2.6707	2000	0.4260	50.2578
0.7665	4.0053	3000	0.3673	45.0413
0.6477	5.3407	4000	0.3415	42.3962
0.581	6.6760	5000	0.3279	40.6422
0.5185	8.0107	6000	0.3181	39.6604
0.4829	9.3460	7000	0.3129	39.2135
0.4284	10.6814	8000	0.3112	38.4829
0.4236	12.0160	9000	0.3074	37.8909
0.4197	13.3514	10000	0.3071	38.1603
0.3905	14.6867	11000	0.3074	37.6797

Framework versions

• Transformers 4.47.1
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.21.0