donut_experiment_bayesian_trial_17

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: 0.4635
• Bleu: 0.0675
• Precisions: [0.8301886792452831, 0.7738095238095238, 0.7272727272727273, 0.6895424836601307]
• Brevity Penalty: 0.0895
• Length Ratio: 0.2930
• Translation Length: 477
• Reference Length: 1628
• Cer: 0.7603
• Wer: 0.8297

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: 0.00018015728878154226
• 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
0.8044	1.0	253	0.7112	0.0610	[0.7535641547861507, 0.6497695852534562, 0.5809018567639257, 0.5125]	0.0987	0.3016	491	1628	0.7647	0.8548
0.3513	2.0	506	0.5640	0.0632	[0.7908902691511387, 0.7089201877934272, 0.6449864498644986, 0.5801282051282052]	0.0934	0.2967	483	1628	0.7549	0.8416
0.2101	3.0	759	0.4754	0.0666	[0.8198757763975155, 0.744131455399061, 0.6802168021680217, 0.6217948717948718]	0.0934	0.2967	483	1628	0.7508	0.8282
0.0756	4.0	1012	0.4635	0.0675	[0.8301886792452831, 0.7738095238095238, 0.7272727272727273, 0.6895424836601307]	0.0895	0.2930	477	1628	0.7603	0.8297

Framework versions

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