whisper-20hrs-class-seed2025

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: 0.3266
• Wer: 14.8052

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: 52
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5645	0.2955	52	0.5506	27.6043
0.414	0.5909	104	0.3880	18.1199
0.3751	0.8864	156	0.3633	17.1872
0.3606	1.1818	208	0.3506	15.9426
0.333	1.4773	260	0.3428	15.9359
0.3534	1.7727	312	0.3347	15.7413
0.2585	2.0682	364	0.3311	14.9025
0.2333	2.3636	416	0.3298	14.8120
0.2845	2.6591	468	0.3280	14.8052
0.3137	2.9545	520	0.3266	14.8052

Framework versions

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