DALPA_GER_3kSource_4kTarget

This model is a fine-tuned version of facebook/nougat-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6322

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 6
• total_train_batch_size: 48
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss
8.5837	1.0	125	0.9670
6.6069	2.0	250	0.8394
5.1117	3.0	375	0.7499
4.1608	4.0	500	0.6896
3.8712	5.0	625	0.6606
3.4876	6.0	750	0.6414
3.6069	7.0	875	0.6391
3.4865	8.0	1000	0.6345
3.2799	9.0	1125	0.6338
3.5958	10.0	1250	0.6324
3.3628	11.0	1375	0.6327
3.7034	12.0	1500	0.6327
3.3937	13.0	1625	0.6328
3.748	14.0	1750	0.6321
3.3068	15.0	1875	0.6322

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 4.1.0
• Tokenizers 0.21.0