layoutlmv3-finetuned-cord_200

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

• Loss: 0.4529
• Precision: 0.9034
• Recall: 0.9169
• F1: 0.9101
• Accuracy: 0.9121

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: 5
• eval_batch_size: 5
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	6.25	250	1.0785	0.6815	0.7575	0.7175	0.7780
1.3902	12.5	500	0.5871	0.8542	0.8683	0.8612	0.8604
1.3902	18.75	750	0.4572	0.8728	0.8937	0.8831	0.8905
0.298	25.0	1000	0.3947	0.8936	0.9117	0.9026	0.9092
0.298	31.25	1250	0.3925	0.8982	0.9177	0.9078	0.9117
0.1023	37.5	1500	0.4290	0.8908	0.9102	0.9004	0.9041
0.1023	43.75	1750	0.4220	0.8980	0.9162	0.9070	0.9117
0.0475	50.0	2000	0.4755	0.8944	0.9064	0.9004	0.8990
0.0475	56.25	2250	0.4635	0.8992	0.9147	0.9069	0.9070
0.0296	62.5	2500	0.4475	0.9019	0.9154	0.9086	0.9117
0.0296	68.75	2750	0.4484	0.9004	0.9139	0.9071	0.9079
0.0242	75.0	3000	0.4529	0.9034	0.9169	0.9101	0.9121

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1