miosipof/whisper-tiny-ft-balbus-sep28k-v1.1

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

• Loss: 0.5089
• Accuracy: 0.7526

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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: 0.5
• training_steps: 800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.692	0.2506	100	0.6725	0.5685
0.6335	0.5013	200	0.6096	0.6623
0.5962	0.7519	300	0.5656	0.7092
0.5685	1.0025	400	0.5457	0.7250
0.5145	1.2531	500	0.5304	0.7336
0.5092	1.5038	600	0.5377	0.7324
0.5035	1.7544	700	0.5133	0.7457
0.4715	2.0050	800	0.5089	0.7526

Framework versions

• Transformers 4.48.0
• Pytorch 2.2.0
• Datasets 3.2.0
• Tokenizers 0.21.0