ThangaTharun/finetunedsmall

This model is a fine-tuned version of distil-whisper/distil-small.en on the ThangaTharun/Barishka2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0136
• Wer: 5.4795

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1
• training_steps: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.1327	1.25	5	2.6560	47.9452
1.7171	2.5	10	1.7472	26.0274
1.0789	3.75	15	0.9175	12.3288
0.0999	5.0	20	0.1663	13.6986
0.0173	6.25	25	0.0855	13.6986
0.004	7.5	30	0.0366	9.5890
0.0017	8.75	35	0.0216	6.8493
0.0008	10.0	40	0.0165	6.8493
0.0005	11.25	45	0.0146	5.4795
0.0005	12.5	50	0.0138	5.4795
0.0005	13.75	55	0.0136	5.4795
0.0004	15.0	60	0.0136	5.4795

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0