whisper-big-krn

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: 0.0000
• Wer: 33.8223

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5092	1.2945	200	0.0978	23.7942
0.0606	2.5890	400	0.0162	25.4019
0.0387	3.8835	600	0.0565	26.5876
0.0227	5.1780	800	0.0097	19.9558
0.0172	6.4725	1000	0.0046	22.3473
0.0157	7.7670	1200	0.0022	29.9638
0.0084	9.0615	1400	0.0024	26.3666
0.0179	10.3560	1600	0.0016	29.0193
0.0089	11.6505	1800	0.0001	28.0145
0.0049	12.9450	2000	0.0054	27.4317
0.0047	14.2395	2200	0.0050	32.1342
0.0041	15.5340	2400	0.0000	34.6664
0.0018	16.8285	2600	0.0001	34.1841
0.0023	18.1230	2800	0.0001	33.8424
0.0024	19.4175	3000	0.0017	31.8931
0.0028	20.7120	3200	0.0000	36.1334
0.0004	22.0065	3400	0.0000	33.6214
0.0	23.3010	3600	0.0000	33.7420
0.0	24.5955	3800	0.0000	33.5209
0.0	25.8900	4000	0.0000	33.4204
0.0	27.1845	4200	0.0000	33.4606
0.0	28.4790	4400	0.0000	33.4606
0.0	29.7735	4600	0.0000	33.8223

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1