whisper-new-nnat-20h-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: 0.4649
• Wer: 20.2922

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: 39
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2486	0.2977	39	0.9010	43.4272
0.6896	0.5954	78	0.5576	23.4393
0.5495	0.8931	117	0.5117	22.0718
0.4965	1.1908	156	0.4941	21.5380
0.4722	1.4885	195	0.4837	21.3975
0.4975	1.7863	234	0.4744	20.6575
0.4619	2.0840	273	0.4700	20.4187
0.4477	2.3817	312	0.4685	20.3765
0.4006	2.6794	351	0.4663	20.2407
0.4139	2.9771	390	0.4649	20.2922

Framework versions

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