whisper-new-nnat-5h-maxcos-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: 0.5519
• Wer: 34.7398

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: 63
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9169	0.2972	63	0.7278	41.9660
0.3981	0.5943	126	0.6059	43.3522
0.3374	0.8915	189	0.5844	50.2318
0.2899	1.1887	252	0.5708	43.8440
0.2836	1.4858	315	0.5627	38.2195
0.3081	1.7830	378	0.5542	35.4283
0.2832	2.0802	441	0.5556	36.0184
0.2731	2.3774	504	0.5539	36.5101
0.2471	2.6745	567	0.5524	35.9341
0.2567	2.9717	630	0.5519	34.7398

Framework versions

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