whisper-large-v3-nyagen-combined-42

This model is a fine-tuned version of openai/whisper-large-v3 on the nyagen dataset. It achieves the following results on the evaluation set:

• Loss: 0.3866
• Wer: 0.4107

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4325	0.4935	200	0.5920	0.3963
0.2825	0.9870	400	0.4630	0.3549
0.2047	1.4787	600	0.4346	0.3190
0.1786	1.9722	800	0.4038	0.3424
0.0973	2.4639	1000	0.4064	0.2505
0.1057	2.9574	1200	0.3866	0.4107
0.0424	3.4491	1400	0.4073	0.2392
0.0591	3.9426	1600	0.4015	0.3277
0.0348	4.4343	1800	0.4354	0.2392
0.0273	4.9278	2000	0.4299	0.2165

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0