2024-01-09_one_stage_subgraphs_weighted_txt_vis_conc_6_ramp

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: 1.2598
• Accuracy: 0.7825
• Exit 0 Accuracy: 0.2725
• Exit 1 Accuracy: 0.7725

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	2.6838	0.13	0.1275	0.0675
No log	1.98	33	2.5115	0.2275	0.125	0.1
No log	3.0	50	2.2857	0.3075	0.1525	0.15
No log	3.96	66	2.0132	0.4075	0.1525	0.3225
No log	4.98	83	1.7290	0.5375	0.1525	0.385
No log	6.0	100	1.4748	0.605	0.155	0.4175
No log	6.96	116	1.2413	0.7275	0.155	0.5475
No log	7.98	133	1.1173	0.745	0.1675	0.535
No log	9.0	150	0.9922	0.7525	0.19	0.565
No log	9.96	166	0.9208	0.765	0.2025	0.5525
No log	10.98	183	0.9245	0.7625	0.2075	0.62
No log	12.0	200	0.8917	0.765	0.2025	0.6525
No log	12.96	216	0.8648	0.7625	0.1975	0.6525
No log	13.98	233	0.8759	0.775	0.2075	0.685
No log	15.0	250	0.9405	0.77	0.215	0.6825
No log	15.96	266	0.9411	0.7725	0.21	0.6975
No log	16.98	283	0.9830	0.7625	0.2225	0.6925
No log	18.0	300	0.9696	0.775	0.2275	0.735
No log	18.96	316	0.9827	0.78	0.23	0.7375
No log	19.98	333	1.0221	0.78	0.235	0.755
No log	21.0	350	1.0305	0.77	0.2225	0.7475
No log	21.96	366	1.0608	0.775	0.2125	0.7325
No log	22.98	383	1.0518	0.7825	0.235	0.7575
No log	24.0	400	1.0250	0.7875	0.24	0.76
No log	24.96	416	1.0568	0.79	0.2425	0.7675
No log	25.98	433	1.0601	0.7875	0.265	0.765
No log	27.0	450	1.0849	0.7925	0.25	0.7625
No log	27.96	466	1.0939	0.78	0.2675	0.7625
No log	28.98	483	1.1370	0.7825	0.265	0.77
0.8593	30.0	500	1.1109	0.785	0.26	0.7575
0.8593	30.96	516	1.1332	0.7825	0.2675	0.7625
0.8593	31.98	533	1.1221	0.7875	0.2675	0.7725
0.8593	33.0	550	1.1503	0.7775	0.265	0.7575
0.8593	33.96	566	1.1640	0.7825	0.265	0.77
0.8593	34.98	583	1.1430	0.79	0.2625	0.7675
0.8593	36.0	600	1.1700	0.7825	0.265	0.775
0.8593	36.96	616	1.1895	0.7775	0.2625	0.755
0.8593	37.98	633	1.1890	0.7825	0.27	0.77
0.8593	39.0	650	1.1726	0.7925	0.2675	0.76
0.8593	39.96	666	1.2200	0.775	0.2725	0.7725
0.8593	40.98	683	1.1801	0.785	0.2725	0.76
0.8593	42.0	700	1.2199	0.7825	0.27	0.77
0.8593	42.96	716	1.2134	0.785	0.27	0.765
0.8593	43.98	733	1.2167	0.78	0.275	0.77
0.8593	45.0	750	1.2335	0.78	0.27	0.7625
0.8593	45.96	766	1.2210	0.785	0.2775	0.7725
0.8593	46.98	783	1.2367	0.7825	0.275	0.77
0.8593	48.0	800	1.2342	0.78	0.27	0.77
0.8593	48.96	816	1.2450	0.7825	0.275	0.7675
0.8593	49.98	833	1.2485	0.785	0.2775	0.7625
0.8593	51.0	850	1.2715	0.78	0.275	0.7675
0.8593	51.96	866	1.2942	0.78	0.2775	0.7675
0.8593	52.98	883	1.2706	0.7825	0.2725	0.7625
0.8593	54.0	900	1.2566	0.78	0.275	0.765
0.8593	54.96	916	1.2581	0.78	0.2725	0.775
0.8593	55.98	933	1.2589	0.7825	0.2725	0.775
0.8593	57.0	950	1.2596	0.7825	0.2725	0.7725
0.8593	57.6	960	1.2598	0.7825	0.2725	0.7725

Framework versions

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