donut-base-sroie-metrics-combined

This model is a fine-tuned version of naver-clova-ix/donut-base on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.2827
• Bleu score: 0.0762
• Precisions: [0.8396396396396396, 0.7886178861788617, 0.7435897435897436, 0.7049180327868853]
• Brevity penalty: 0.0993
• Length ratio: 0.3021
• Translation length: 555
• Reference length: 1837
• Cer: 0.7452
• Wer: 0.8162
• Cer Hugging Face: 0.7544
• Wer Hugging Face: 0.8233

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: 2e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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 score	Precisions	Brevity penalty	Length ratio	Translation length	Reference length	Cer	Wer	Cer Hugging Face	Wer Hugging Face
No log	0.99	62	0.3478	0.0756	[0.8178571428571428, 0.7625754527162978, 0.7142857142857143, 0.6711590296495957]	0.1022	0.3048	560	1837	0.7474	0.8243	0.7570	0.8333
0.2634	2.0	125	0.2873	0.0763	[0.8345323741007195, 0.7829614604462475, 0.7418604651162791, 0.7029972752043597]	0.0999	0.3027	556	1837	0.7435	0.8219	0.7527	0.8288
0.2634	2.99	187	0.2817	0.0777	[0.8369175627240143, 0.7838383838383839, 0.7476851851851852, 0.7127371273712737]	0.1011	0.3038	558	1837	0.7407	0.8152	0.7498	0.8215
0.263	3.97	248	0.2827	0.0762	[0.8396396396396396, 0.7886178861788617, 0.7435897435897436, 0.7049180327868853]	0.0993	0.3021	555	1837	0.7452	0.8162	0.7544	0.8233

Framework versions

• Transformers 4.40.0.dev0
• Pytorch 2.1.0
• Datasets 2.18.0
• Tokenizers 0.15.2