donut_experiment_bayesian_trial_19

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: 0.5754
• Bleu: 0.0724
• Precisions: [0.8450413223140496, 0.7892271662763466, 0.7486486486486487, 0.7028753993610224]
• Brevity Penalty: 0.0941
• Length Ratio: 0.2973
• Translation Length: 484
• Reference Length: 1628
• Cer: 0.7493
• Wer: 0.8177

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: 1.0668629620167924e-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: 3
• 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
0.0069	1.0	253	0.5825	0.0710	[0.8423236514522822, 0.7858823529411765, 0.7418478260869565, 0.6977491961414791]	0.0928	0.2961	482	1628	0.7509	0.8197
0.0113	2.0	506	0.5684	0.0703	[0.841995841995842, 0.785377358490566, 0.7411444141689373, 0.6935483870967742]	0.0921	0.2955	481	1628	0.7505	0.8199
0.0074	3.0	759	0.5754	0.0724	[0.8450413223140496, 0.7892271662763466, 0.7486486486486487, 0.7028753993610224]	0.0941	0.2973	484	1628	0.7493	0.8177

Framework versions

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