whisper-5hrs-meta-seed333

This model is a fine-tuned version of openai/whisper-large-v2 on the JASMIN-CGN dataset. It achieves the following results on the evaluation set:

• Loss: 0.4032
• Wer: 18.9888

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: 0.0001
• train_batch_size: 48
• eval_batch_size: 32
• 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: 12
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9872	0.2927	12	1.1235	36.6189
0.8638	0.5854	24	0.8173	34.8106
0.5604	0.8780	36	0.5658	26.0509
0.444	1.1707	48	0.4771	19.8712
0.432	1.4634	60	0.4417	18.9553
0.3875	1.7561	72	0.4236	19.6397
0.3616	2.0488	84	0.4130	20.2872
0.3057	2.3415	96	0.4078	20.2402
0.3899	2.6341	108	0.4046	19.0392
0.3609	2.9268	120	0.4032	18.9888

Framework versions

• PEFT 0.16.0
• Transformers 4.57.3
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.22.2