miosipof/asr_michael_medium_v3.2b_augmented

Whisper Medium

This model is a fine-tuned version of openai/whisper-medium on the b-brave/speech_disorders_voice dataset. It achieves the following results on the evaluation set:

• Loss: 0.3733
• Wer: 18.4644

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: 0.001
• train_batch_size: 128
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 16
• training_steps: 64
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.0681	3.7647	16	1.7949	48.6289
0.5772	7.5294	32	0.3912	22.3035
0.1166	11.2941	48	0.3734	18.6472
0.0608	15.0588	64	0.3733	18.4644

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.2.0
• Datasets 2.21.0
• Tokenizers 0.19.1