donut_experiment_bayesian_trial_20

This model is a fine-tuned version of naver-clova-ix/donut-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6042
• Bleu: 0.0711
• Precisions: [0.8440748440748441, 0.7924528301886793, 0.7493188010899182, 0.7064516129032258]
• Brevity Penalty: 0.0921
• Length Ratio: 0.2955
• Translation Length: 481
• Reference Length: 1628
• Cer: 0.7518
• Wer: 0.8208

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: 1.4151037707088747e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Precisions	Brevity Penalty	Length Ratio	Translation Length	Reference Length	Cer	Wer
0.0012	1.0	253	0.6392	0.0724	[0.8333333333333334, 0.7762237762237763, 0.7365591397849462, 0.6952380952380952]	0.0954	0.2985	486	1628	0.7482	0.8193
0.0068	2.0	506	0.6212	0.0697	[0.8413361169102297, 0.7867298578199052, 0.7452054794520548, 0.7012987012987013]	0.0908	0.2942	479	1628	0.7527	0.8230
0.0087	3.0	759	0.6105	0.0687	[0.83125, 0.7706855791962175, 0.726775956284153, 0.6828478964401294]	0.0915	0.2948	480	1628	0.7573	0.8282
0.0056	4.0	1012	0.6042	0.0711	[0.8440748440748441, 0.7924528301886793, 0.7493188010899182, 0.7064516129032258]	0.0921	0.2955	481	1628	0.7518	0.8208

Framework versions

• Transformers 4.40.0
• Pytorch 2.1.0
• Datasets 2.18.0
• Tokenizers 0.19.1