sxandie/doc-ai-information-extraction

sougemi_model

This model is a fine-tuned version of microsoft/layoutlmv3-base on the dataset created. It achieves the following results on the evaluation set:

• Loss: 0.1812
• Precision: 0.8454
• Recall: 0.8913
• F1: 0.8677
• Accuracy: 0.9534

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	33.33	100	0.7803	0.8966	0.8478	0.8715	0.9663
No log	66.67	200	0.3016	0.8696	0.8696	0.8696	0.9767
No log	100.0	300	0.1623	0.9130	0.9130	0.9130	0.9819
No log	133.33	400	0.1680	0.8454	0.8913	0.8677	0.9637
0.5801	166.67	500	0.1812	0.8454	0.8913	0.8677	0.9534
0.5801	200.0	600	0.1231	0.8947	0.9239	0.9091	0.9715
0.5801	233.33	700	0.1363	0.8617	0.8804	0.8710	0.9663
0.5801	266.67	800	0.1949	0.8333	0.8696	0.8511	0.9508
0.5801	300.0	900	0.1749	0.8163	0.8696	0.8421	0.9534
0.0607	333.33	1000	0.1817	0.8163	0.8696	0.8421	0.9534

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2