VoiceMath-Tiny

This model is a fine-tuned version of openai/whisper-tiny.en on the model_eng dataset. It achieves the following results on the evaluation set:

• Loss: 0.0162
• Wer: 9.4429

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: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	15	0.2506	15.9585
No log	2.0	30	0.1447	13.9754
No log	3.0	45	0.0945	11.4259
No log	4.0	60	0.0657	11.7092
No log	5.0	75	0.0480	11.4259
No log	6.0	90	0.0322	9.9150
No log	7.0	105	0.0243	10.1983
No log	8.0	120	0.0196	9.2540
No log	9.0	135	0.0171	9.4429
No log	10.0	150	0.0162	9.4429

Framework versions

• Transformers 4.30.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.13.3