whisper-5hrs-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: 0.3990
• Wer: 18.3380

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: 13
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0653	0.2955	13	1.1094	36.3304
0.8018	0.5909	26	0.7474	33.8578
0.5938	0.8864	39	0.5354	24.9304
0.4762	1.1818	52	0.4591	21.4849
0.3695	1.4773	65	0.4315	19.2304
0.4079	1.7727	78	0.4162	19.3780
0.3828	2.0682	91	0.4078	18.7506
0.3489	2.3636	104	0.4036	18.4789
0.3686	2.6591	117	0.4006	18.3145
0.3816	2.9545	130	0.3990	18.3380

Framework versions

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