2024-01-03_one_stage_subgraphs_weighted_txt_vis_conc_6_gate

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

• Loss: 1.2712
• Accuracy: 0.765
• Exit 0 Accuracy: 0.0675
• Exit 1 Accuracy: 0.13

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: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 24
• total_train_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 60

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Exit 0 Accuracy	Exit 1 Accuracy
No log	0.96	16	2.6688	0.16	0.0525	0.08
No log	1.98	33	2.4616	0.23	0.055	0.0525
No log	3.0	50	2.2105	0.3525	0.0575	0.0575
No log	3.96	66	1.9883	0.42	0.0575	0.0525
No log	4.98	83	1.7863	0.525	0.0625	0.0575
No log	6.0	100	1.4980	0.6125	0.0675	0.04
No log	6.96	116	1.3248	0.6475	0.0675	0.055
No log	7.98	133	1.1715	0.6875	0.065	0.0625
No log	9.0	150	1.0884	0.6975	0.0675	0.0625
No log	9.96	166	1.0221	0.725	0.0675	0.0525
No log	10.98	183	0.9646	0.7375	0.0675	0.0475
No log	12.0	200	0.9562	0.75	0.065	0.0525
No log	12.96	216	0.8957	0.75	0.065	0.035
No log	13.98	233	0.9117	0.76	0.065	0.055
No log	15.0	250	0.8972	0.765	0.065	0.0375
No log	15.96	266	0.9015	0.765	0.065	0.05
No log	16.98	283	0.9712	0.7625	0.0675	0.04
No log	18.0	300	0.9805	0.755	0.0675	0.0825
No log	18.96	316	0.9794	0.7525	0.0675	0.05
No log	19.98	333	1.0191	0.75	0.0675	0.0575
No log	21.0	350	1.0427	0.745	0.0675	0.0725
No log	21.96	366	0.9744	0.77	0.065	0.0925
No log	22.98	383	1.0432	0.7575	0.065	0.115
No log	24.0	400	1.0682	0.7625	0.065	0.105
No log	24.96	416	1.0981	0.7675	0.0675	0.1175
No log	25.98	433	1.1199	0.765	0.0675	0.1075
No log	27.0	450	1.1305	0.76	0.0675	0.1075
No log	27.96	466	1.1391	0.7625	0.0675	0.1125
No log	28.98	483	1.1646	0.765	0.0675	0.095
0.3865	30.0	500	1.1655	0.7625	0.0675	0.0975
0.3865	30.96	516	1.1787	0.75	0.0675	0.1025
0.3865	31.98	533	1.1661	0.7725	0.0675	0.11
0.3865	33.0	550	1.1744	0.7725	0.0675	0.11
0.3865	33.96	566	1.2073	0.77	0.0675	0.095
0.3865	34.98	583	1.2425	0.75	0.0675	0.09
0.3865	36.0	600	1.2566	0.7525	0.0675	0.0825
0.3865	36.96	616	1.2562	0.7525	0.0675	0.085
0.3865	37.98	633	1.2366	0.75	0.0675	0.0825
0.3865	39.0	650	1.2024	0.77	0.0675	0.0825
0.3865	39.96	666	1.2182	0.7675	0.0675	0.09
0.3865	40.98	683	1.2355	0.7575	0.0675	0.0825
0.3865	42.0	700	1.2351	0.765	0.0675	0.09
0.3865	42.96	716	1.2479	0.7575	0.0675	0.1025
0.3865	43.98	733	1.2311	0.7675	0.0675	0.105
0.3865	45.0	750	1.2517	0.765	0.0675	0.0975
0.3865	45.96	766	1.2442	0.7675	0.0675	0.1025
0.3865	46.98	783	1.2380	0.765	0.0675	0.11
0.3865	48.0	800	1.2502	0.77	0.0675	0.1025
0.3865	48.96	816	1.2488	0.77	0.0675	0.1025
0.3865	49.98	833	1.2498	0.77	0.0675	0.105
0.3865	51.0	850	1.2554	0.7725	0.0675	0.12
0.3865	51.96	866	1.2683	0.7625	0.0675	0.1225
0.3865	52.98	883	1.2607	0.7675	0.0675	0.1325
0.3865	54.0	900	1.2689	0.765	0.0675	0.13
0.3865	54.96	916	1.2597	0.765	0.0675	0.13
0.3865	55.98	933	1.2672	0.7675	0.0675	0.125
0.3865	57.0	950	1.2714	0.765	0.0675	0.13
0.3865	57.6	960	1.2712	0.765	0.0675	0.13

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3