donut_experiment_2

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: 0.3855
• Bleu: 0.0663
• Precisions: [0.8273684210526315, 0.7703349282296651, 0.7285318559556787, 0.6842105263157895]
• Brevity Penalty: 0.0883
• Length Ratio: 0.2918
• Translation Length: 475
• Reference Length: 1628
• Cer: 0.7539
• Wer: 0.8251

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: 2e-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
0.9517	1.0	253	0.5797	0.0543	[0.7160751565762005, 0.6184834123222749, 0.5671232876712329, 0.5097402597402597]	0.0908	0.2942	479	1628	0.7738	0.8500
0.3907	2.0	506	0.4532	0.0590	[0.7851063829787234, 0.711864406779661, 0.6657303370786517, 0.6220735785953178]	0.0851	0.2887	470	1628	0.7610	0.8370
0.3245	3.0	759	0.4102	0.0625	[0.8008474576271186, 0.7397590361445783, 0.7011173184357542, 0.6611295681063123]	0.0864	0.2899	472	1628	0.7593	0.8336
0.2318	4.0	1012	0.3855	0.0663	[0.8273684210526315, 0.7703349282296651, 0.7285318559556787, 0.6842105263157895]	0.0883	0.2918	475	1628	0.7539	0.8251

Framework versions

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