whisper-5hrs-meta-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: 0.4036
• Wer: 19.4451

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
1.023	0.2927	12	1.1245	36.7229
0.8869	0.5854	24	0.8261	35.1931
0.581	0.8780	36	0.5729	26.3160
0.4777	1.1707	48	0.4823	19.4485
0.3935	1.4634	60	0.4439	21.5352
0.3731	1.7561	72	0.4245	20.3912
0.4175	2.0488	84	0.4144	19.9181
0.3771	2.3415	96	0.4087	19.4283
0.3764	2.6341	108	0.4049	19.3176
0.4073	2.9268	120	0.4036	19.4451

Framework versions

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