whisper-large-v3-DODa2-AT

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2468
• Wer: 56.5801

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 150
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	120	1.4419	82.2792
No log	2.0	240	1.1389	63.8038
No log	3.0	360	1.0669	70.8944
No log	4.0	480	1.0693	58.7439
1.1787	5.0	600	1.1195	60.7856
1.1787	6.0	720	1.1659	60.8411
1.1787	7.0	840	1.1861	58.1780
1.1787	8.0	960	1.2127	59.7980
0.0813	9.0	1080	1.2412	57.2015
0.0813	10.0	1200	1.2468	56.5801

Framework versions

• Transformers 4.46.2
• Pytorch 2.4.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.0