whisper-new-nnat-10h-random-seed333

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: 0.5103
• Wer: 23.6922

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: 21
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9189	0.2877	21	1.2450	43.9002
0.6101	0.5753	42	0.7081	32.2812
0.4723	0.8630	63	0.5897	25.8605
0.444	1.1507	84	0.5524	23.8140
0.3691	1.4384	105	0.5373	23.1677
0.3649	1.7260	126	0.5261	23.5424
0.3777	2.0137	147	0.5180	23.1396
0.3763	2.3014	168	0.5140	24.1793
0.3635	2.5890	189	0.5112	25.0129
0.4017	2.8767	210	0.5103	23.6922

Framework versions

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