donut_experiment_bayesian_trial_7

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: 0.3786
• Bleu: 0.0669
• Precisions: [0.8477801268498943, 0.7836538461538461, 0.7465181058495822, 0.7052980132450332]
• Brevity Penalty: 0.0870
• Length Ratio: 0.2905
• Translation Length: 473
• Reference Length: 1628
• Cer: 0.7532
• Wer: 0.8192

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: 3.540464175534869e-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: 5
• 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
0.5323	1.0	253	0.4204	0.0580	[0.7710084033613446, 0.6778042959427207, 0.6132596685082873, 0.5639344262295082]	0.0889	0.2924	476	1628	0.7617	0.8431
0.2487	2.0	506	0.3788	0.0609	[0.8123667377398721, 0.7402912621359223, 0.6929577464788732, 0.6476510067114094]	0.0845	0.2881	469	1628	0.7561	0.8279
0.1746	3.0	759	0.3551	0.0652	[0.836864406779661, 0.7759036144578313, 0.729050279329609, 0.6843853820598007]	0.0864	0.2899	472	1628	0.7541	0.8213
0.1191	4.0	1012	0.3690	0.0680	[0.8547368421052631, 0.784688995215311, 0.7451523545706371, 0.7039473684210527]	0.0883	0.2918	475	1628	0.7514	0.8192
0.1072	5.0	1265	0.3786	0.0669	[0.8477801268498943, 0.7836538461538461, 0.7465181058495822, 0.7052980132450332]	0.0870	0.2905	473	1628	0.7532	0.8192

Framework versions

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