whisper-5hrs-class-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: 0.4019
• Wer: 18.7305

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
1.0474	0.2955	13	1.1135	36.5921
0.7895	0.5909	26	0.7599	36.0385
0.5323	0.8864	39	0.5341	24.2493
0.4659	1.1818	52	0.4599	20.8139
0.4317	1.4773	65	0.4340	19.8510
0.4217	1.7727	78	0.4181	19.4619
0.4044	2.0682	91	0.4103	18.8211
0.3859	2.3636	104	0.4055	18.6064
0.3755	2.6591	117	0.4029	18.7473
0.365	2.9545	130	0.4019	18.7305

Framework versions

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