whisper-new-nnat-5h-caps-ecapa

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: 0.5138
• Wer: 23.9638

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 adamw_torch_fused 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: 32
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2318	0.2991	32	0.9904	46.0029
0.5262	0.5981	64	0.6141	29.3120
0.4205	0.8972	96	0.5630	26.0151
0.4261	1.1963	128	0.5430	24.1886
0.3874	1.4953	160	0.5303	23.8702
0.3709	1.7944	192	0.5222	25.4063
0.3507	2.0935	224	0.5185	25.0925
0.3254	2.3925	256	0.5164	24.5727
0.3059	2.6916	288	0.5143	23.5283
0.3325	2.9907	320	0.5138	23.9638

Framework versions

• PEFT 0.17.1
• Transformers 4.57.6
• Pytorch 2.8.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2