whisper-large-nya

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

• Loss: 0.4712
• Wer: 21.5239

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: 2.5e-05
• train_batch_size: 8
• eval_batch_size: 4
• 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: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2416	0.99	500	0.5146	34.7076
0.1343	1.97	1000	0.4138	28.1748
0.0792	2.96	1500	0.4268	31.3290
0.0372	3.94	2000	0.4256	32.8057
0.0246	4.93	2500	0.4354	22.0673
0.0097	5.92	3000	0.4532	25.1742
0.003	6.9	3500	0.4595	21.0396
0.0005	7.89	4000	0.4586	21.3113
0.0007	8.87	4500	0.4653	21.7129
0.0002	9.86	5000	0.4712	21.5239

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2