UDOP-finetuned-DocLayNet-3

This model is a fine-tuned version of on an unknown dataset. It achieves the following results on the evaluation set:

• eval_loss: 0.7407
• eval_precision: 0.6058
• eval_recall: 0.5870
• eval_f1: 0.5962
• eval_accuracy: 0.7863
• eval_runtime: 16.2128
• eval_samples_per_second: 3.886
• eval_steps_per_second: 1.974
• epoch: 18.6
• step: 800

Training procedure

Training code:

  https://github.com/mit1280/Document-AI/blob/main/UDOPEncoderModel_fine_tune_DocLayNet.ipynb

Inference code:

  https://github.com/mit1280/Document-AI/blob/main/UDOP_DocLayNet_Inference.ipynb

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 1000

Framework versions

• Transformers 4.39.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2