donut-base-sroie-bayesian-optimization

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: 0.1396
• Bleu: 0.0196
• Precisions: [0.9883177570093458, 0.9724655819774718, 0.954177897574124, 0.9328467153284672]
• Brevity Penalty: 0.0203
• Length Ratio: 0.2043
• Translation Length: 856
• Reference Length: 4190
• Cer: 0.8584
• Wer: 1.0

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: 1.2010406976282324e-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: 5
• 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
0.021	1.0	253	0.1656	0.0194	[0.9848130841121495, 0.9649561952440551, 0.9420485175202157, 0.9153284671532846]	0.0203	0.2043	856	4190	0.8596	1.0
0.0353	2.0	506	0.1501	0.0195	[0.9813736903376019, 0.9588528678304239, 0.9328859060402684, 0.9026162790697675]	0.0207	0.2050	859	4190	0.8595	1.0
0.0417	3.0	759	0.1423	0.0195	[0.9871495327102804, 0.9699624530663329, 0.9501347708894878, 0.927007299270073]	0.0203	0.2043	856	4190	0.8586	1.0
0.0308	4.0	1012	0.1403	0.0193	[0.9859649122807017, 0.9674185463659147, 0.9460188933873145, 0.9210526315789473]	0.0202	0.2041	855	4190	0.8593	1.0
0.0464	5.0	1265	0.1396	0.0196	[0.9883177570093458, 0.9724655819774718, 0.954177897574124, 0.9328467153284672]	0.0203	0.2043	856	4190	0.8584	1.0

Framework versions

• Transformers 4.41.0.dev0
• Pytorch 2.1.0
• Datasets 2.19.0
• Tokenizers 0.19.1