whisper-small-mg

This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6241
• Wer: 0.4037

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
• 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_steps: 200
• training_steps: 2109
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.7177	0.3556	250	0.9988	0.5509
0.8811	0.7112	500	0.7718	0.4624
0.7301	1.0669	750	0.7010	0.4345
0.6074	1.4225	1000	0.6691	0.4246
0.5995	1.7781	1250	0.6460	0.4179
0.5464	2.1337	1500	0.6365	0.4200
0.4915	2.4893	1750	0.6286	0.4251
0.4787	2.8450	2000	0.6241	0.4037

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.3.1
• Tokenizers 0.21.0