whisper-5hrs-meta-seed42

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: 0.4002
• Wer: 19.8007

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
0.9996	0.3095	13	1.1125	36.3438
0.7769	0.6190	26	0.7698	34.4818
0.5467	0.9286	39	0.5419	25.0646
0.4021	1.2381	52	0.4640	20.7065
0.4157	1.5476	65	0.4331	20.1932
0.4022	1.8571	78	0.4167	20.1161
0.3593	2.1667	91	0.4075	19.8913
0.3637	2.4762	104	0.4027	19.1834
0.3847	2.7857	117	0.4002	19.8007

Framework versions

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