model_whisper_2

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: 0.2258
• Char Accuracy: 0.9893

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.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_steps: 20
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Char Accuracy
0.5294	0.4706	5	0.2149	0.9896
0.6061	0.9412	10	0.2163	0.9891
0.3241	1.3765	15	0.2167	0.9890
0.4054	1.8471	20	0.2336	0.9907
0.2066	2.2824	25	0.2302	0.9887
0.1746	2.7529	30	0.2275	0.9887
0.1746	3.0	33	0.2258	0.9893

Framework versions

• Transformers 5.9.0
• Pytorch 2.12.0+cu126
• Datasets 2.18.0
• Tokenizers 0.22.2