whisper_CN_2

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: 0.2958
• Wer: 223.4445

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: 5e-06
• train_batch_size: 8
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0378	2.84	2000	0.1870	195.0945
0.0023	5.67	4000	0.2071	240.6880
0.0006	8.51	6000	0.2149	160.2463
0.0003	11.35	8000	0.2226	164.5148
0.0002	28.4	10000	0.2332	160.5649
0.0002	34.08	12000	0.2460	301.9112
0.0001	39.76	14000	0.2598	266.9357
0.0	45.45	16000	0.2776	165.8526
0.0	51.12	18000	0.2896	204.3321
0.0	56.79	20000	0.2958	223.4445

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.1+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0