donut_experiment_bayesian_trial_11

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: 0.4496
• Bleu: 0.0662
• Precisions: [0.8162839248434238, 0.7440758293838863, 0.7013698630136986, 0.6623376623376623]
• Brevity Penalty: 0.0908
• Length Ratio: 0.2942
• Translation Length: 479
• Reference Length: 1628
• Cer: 0.7615
• Wer: 0.8328

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: 0.00012678733283601488
• 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
0.5513	1.0	253	0.5209	0.0666	[0.7287128712871287, 0.6450892857142857, 0.578005115089514, 0.5269461077844312]	0.1082	0.3102	505	1628	0.7584	0.8461
0.2612	2.0	506	0.5017	0.0648	[0.7962577962577962, 0.7287735849056604, 0.6784741144414169, 0.6225806451612903]	0.0921	0.2955	481	1628	0.7542	0.8349
0.1638	3.0	759	0.4666	0.0615	[0.7995824634655533, 0.6990521327014217, 0.6356164383561644, 0.5909090909090909]	0.0908	0.2942	479	1628	0.7636	0.8392
0.059	4.0	1012	0.4496	0.0662	[0.8162839248434238, 0.7440758293838863, 0.7013698630136986, 0.6623376623376623]	0.0908	0.2942	479	1628	0.7615	0.8328

Framework versions

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