donut_experiment_bayesian_trial_15

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: 0.5777
• Bleu: 0.0659
• Precisions: [0.8158995815899581, 0.7434679334916865, 0.7060439560439561, 0.6644951140065146]
• Brevity Penalty: 0.0902
• Length Ratio: 0.2936
• Translation Length: 478
• Reference Length: 1628
• Cer: 0.7557
• Wer: 0.8239

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: 2.349414650597281e-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
0.0066	1.0	253	0.5790	0.0648	[0.8305084745762712, 0.7686746987951807, 0.7262569832402235, 0.6843853820598007]	0.0864	0.2899	472	1628	0.7593	0.8258
0.0143	2.0	506	0.5824	0.0663	[0.8225469728601252, 0.7511848341232228, 0.7041095890410959, 0.6525974025974026]	0.0908	0.2942	479	1628	0.7577	0.8265
0.009	3.0	759	0.5826	0.0640	[0.8185654008438819, 0.7458033573141487, 0.7055555555555556, 0.6600660066006601]	0.0876	0.2912	474	1628	0.7553	0.8248
0.0103	4.0	1012	0.5777	0.0659	[0.8158995815899581, 0.7434679334916865, 0.7060439560439561, 0.6644951140065146]	0.0902	0.2936	478	1628	0.7557	0.8239

Framework versions

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