whisper-new-nnat-20h-random-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: 0.4691
• Wer: 20.6950

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: 45
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8351	0.3	45	0.8048	39.6806
0.4512	0.6	90	0.5502	24.4649
0.4241	0.9	135	0.5124	23.5611
0.3697	1.2	180	0.4979	22.1374
0.3582	1.5	225	0.4848	21.3928
0.363	1.8	270	0.4774	21.5942
0.3568	2.1	315	0.4744	20.7886
0.3357	2.4	360	0.4710	20.6669
0.3389	2.7	405	0.4697	20.8402
0.338	3.0	450	0.4691	20.6950

Framework versions

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