donut_experiment_bayesian_trial_10

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: 0.4219
• Bleu: 0.0632
• Precisions: [0.809322033898305, 0.7493975903614458, 0.7094972067039106, 0.6644518272425249]
• Brevity Penalty: 0.0864
• Length Ratio: 0.2899
• Translation Length: 472
• Reference Length: 1628
• Cer: 0.7596
• Wer: 0.8312

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: 1.0082458996730595e-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
0.3888	1.0	253	0.5110	0.0673	[0.7909836065573771, 0.7099767981438515, 0.6657754010695187, 0.6277602523659306]	0.0967	0.2998	488	1628	0.7690	0.8412
0.326	2.0	506	0.4539	0.0654	[0.7908902691511387, 0.7276995305164319, 0.6775067750677507, 0.6153846153846154]	0.0934	0.2967	483	1628	0.7604	0.8362
0.3191	3.0	759	0.4256	0.0654	[0.7837837837837838, 0.7287735849056604, 0.6893732970027248, 0.6451612903225806]	0.0921	0.2955	481	1628	0.7599	0.8331
0.2632	4.0	1012	0.4219	0.0632	[0.809322033898305, 0.7493975903614458, 0.7094972067039106, 0.6644518272425249]	0.0864	0.2899	472	1628	0.7596	0.8312

Framework versions

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