_base_nougat_jawi

This model is a fine-tuned version of facebook/nougat-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4371

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: 30

Training results

Training Loss	Epoch	Step	Validation Loss
12.3256	1.0	84	1.9774
11.0152	2.0	168	1.7704
10.0413	3.0	252	1.6441
9.5176	4.0	336	1.5924
8.4833	5.0	420	1.5436
8.6871	6.0	504	1.4766
8.2367	7.0	588	1.4353
7.5666	8.0	672	1.3372
6.3156	9.0	756	1.0868
4.7622	10.0	840	0.8802
4.2315	11.0	924	0.7787
3.6727	12.0	1008	0.6649
2.944	13.0	1092	0.5857
2.9873	14.0	1176	0.5698
2.326	15.0	1260	0.5270
2.0722	16.0	1344	0.5679
2.1138	17.0	1428	0.4879
2.1412	18.0	1512	0.4786
1.9	19.0	1596	0.4855
1.8338	20.0	1680	0.4556
1.8543	21.0	1764	0.4475
1.5969	22.0	1848	0.4443
1.7896	23.0	1932	0.4375
1.5593	24.0	2016	0.4359
1.5125	25.0	2100	0.4303
1.672	26.0	2184	0.4348
1.6588	27.0	2268	0.4370
1.6671	28.0	2352	0.4371

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0