donut-base-sroie-v2

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

• Loss: 0.4355
• Bleu: 0.8879
• Precisions: [0.943646408839779, 0.9119229045271179, 0.8854285064787452, 0.860009225092251]
• Brevity Penalty: 0.9868
• Length Ratio: 0.9869
• Translation Length: 4525
• Reference Length: 4585
• Cer: 0.0857
• Wer: 0.2978

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	Precisions	Brevity Penalty	Length Ratio	Translation Length	Reference Length	Cer	Wer
No log	0.99	62	0.4638	0.8823	[0.9399823477493381, 0.9044528977399866, 0.8772128915115751, 0.8514851485148515]	0.9884	0.9884	4532	4585	0.0912	0.3085
0.0043	2.0	125	0.4421	0.8853	[0.9405155320555189, 0.9059428060768543, 0.8794470881486517, 0.8537931034482759]	0.9899	0.9900	4539	4585	0.0889	0.3050
0.0043	2.99	187	0.4328	0.8904	[0.9399122807017544, 0.9068267734044919, 0.8809201623815968, 0.8558682223747426]	0.9945	0.9945	4560	4585	0.0842	0.2939
0.0106	3.97	248	0.4355	0.8879	[0.943646408839779, 0.9119229045271179, 0.8854285064787452, 0.860009225092251]	0.9868	0.9869	4525	4585	0.0857	0.2978

Framework versions

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