donut_experiment_bayesian_trial_1

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: 0.4856
• Bleu: 0.0696
• Precisions: [0.8274428274428275, 0.7712264150943396, 0.7356948228882834, 0.6935483870967742]
• Brevity Penalty: 0.0921
• Length Ratio: 0.2955
• Translation Length: 481
• Reference Length: 1628
• Cer: 0.7518
• Wer: 0.8247

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: 7.862160042303398e-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: 2
• 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
0.0892	1.0	253	0.5425	0.0656	[0.8221757322175732, 0.7505938242280285, 0.7005494505494505, 0.6482084690553745]	0.0902	0.2936	478	1628	0.7544	0.8316
0.0402	2.0	506	0.4856	0.0696	[0.8274428274428275, 0.7712264150943396, 0.7356948228882834, 0.6935483870967742]	0.0921	0.2955	481	1628	0.7518	0.8247

Framework versions

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