parambharat/whisper-small-kn

Whisper Small Kn - Bharat Ramanathan

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1398
• Wer: 23.8167

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: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4126	0.1	500	2.2797	127.2639
0.2099	0.1	1000	0.1774	28.2494
0.1736	0.2	1500	0.1565	27.5733
0.1506	0.3	2000	0.1514	26.0331
0.1373	0.4	2500	0.1494	24.4177
0.1298	0.5	3000	0.1456	25.0563
0.1198	1.06	3500	0.1436	24.4177
0.1102	0.1	4000	0.1452	24.2675
0.1097	0.2	4500	0.1402	24.3050
0.105	0.3	5000	0.1398	23.8167

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2