TR-Fin-Table-Structure-HoixiFinetuned-v1

This model is a fine-tuned version of microsoft/table-transformer-structure-recognition-v1.1-all on the tr-fin_table-dataset-v4 dataset. It achieves the following results on the evaluation set:

• Loss: 1.5728

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: 5e-06
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
3.3718	12.5	50	1.8326
2.352	25.0	100	1.8074
1.8852	37.5	150	1.7669
2.4453	50.0	200	1.6969
1.7715	62.5	250	1.6691
1.8492	75.0	300	1.6298
1.9541	87.5	350	1.6240
1.6634	100.0	400	1.6063
2.0306	112.5	450	1.5812
2.0784	125.0	500	1.5645
1.7727	137.5	550	1.5720
2.7045	150.0	600	1.5749
2.2544	162.5	650	1.5736
2.06	175.0	700	1.5837
1.8153	187.5	750	1.5631
1.9772	200.0	800	1.5747
2.3578	212.5	850	1.5718
1.5885	225.0	900	1.5746
1.8051	237.5	950	1.5705
2.5822	250.0	1000	1.5728

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu126
• Datasets 3.5.0
• Tokenizers 0.21.0