2024-01-11_one_stage_subgraphs_entropyreg_txt_vis_conc_1_4_8_12_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: 1.5748
• Accuracy: 0.775
• Exit 0 Accuracy: 0.0675
• Exit 1 Accuracy: 0.405
• Exit 2 Accuracy: 0.6925
• Exit 3 Accuracy: 0.7725
• Exit 4 Accuracy: 0.775

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	Exit 2 Accuracy	Exit 3 Accuracy	Exit 4 Accuracy
No log	0.96	16	2.6830	0.1475	0.08	0.07	0.0625	0.0625	0.0925
No log	1.98	33	2.5162	0.2225	0.0875	0.115	0.0625	0.0625	0.155
No log	3.0	50	2.3234	0.3075	0.08	0.125	0.0625	0.0625	0.255
No log	3.96	66	2.0222	0.4525	0.09	0.13	0.0625	0.0625	0.38
No log	4.98	83	1.7524	0.5625	0.085	0.125	0.0625	0.0625	0.5025
No log	6.0	100	1.5038	0.6175	0.08	0.12	0.0625	0.0625	0.5475
No log	6.96	116	1.3022	0.6775	0.085	0.1125	0.0625	0.0625	0.6075
No log	7.98	133	1.1943	0.7	0.0825	0.1225	0.0625	0.0625	0.625
No log	9.0	150	1.0962	0.7275	0.06	0.13	0.0625	0.0625	0.65
No log	9.96	166	0.9921	0.7375	0.07	0.125	0.0625	0.0625	0.695
No log	10.98	183	0.9680	0.735	0.065	0.125	0.0625	0.0625	0.725
No log	12.0	200	1.0108	0.715	0.065	0.12	0.0625	0.0625	0.7225
No log	12.96	216	0.9279	0.745	0.0625	0.1325	0.0625	0.0625	0.7375
No log	13.98	233	0.9503	0.7375	0.0675	0.1175	0.0625	0.0625	0.7475
No log	15.0	250	0.9331	0.755	0.0675	0.13	0.0625	0.0625	0.7675
No log	15.96	266	0.9916	0.755	0.0775	0.13	0.0625	0.0625	0.7525
No log	16.98	283	0.9925	0.7625	0.0775	0.13	0.075	0.0825	0.7675
No log	18.0	300	1.0271	0.765	0.0725	0.1375	0.08	0.1275	0.765
No log	18.96	316	1.0369	0.78	0.0675	0.1375	0.0725	0.2125	0.775
No log	19.98	333	1.1098	0.755	0.06	0.1375	0.08	0.435	0.7525
No log	21.0	350	1.1733	0.745	0.0575	0.13	0.095	0.5625	0.7475
No log	21.96	366	1.1522	0.765	0.055	0.1325	0.11	0.635	0.7625
No log	22.98	383	1.1609	0.7525	0.0575	0.1325	0.1625	0.715	0.745
No log	24.0	400	1.1421	0.76	0.06	0.13	0.28	0.7475	0.76
No log	24.96	416	1.3286	0.74	0.06	0.13	0.295	0.745	0.74
No log	25.98	433	1.2456	0.76	0.0625	0.125	0.33	0.745	0.7575
No log	27.0	450	1.2226	0.78	0.0625	0.1225	0.3775	0.77	0.785
No log	27.96	466	1.2414	0.77	0.0625	0.12	0.42	0.785	0.7725
No log	28.98	483	1.3176	0.77	0.0675	0.1575	0.43	0.7575	0.7625
1.479	30.0	500	1.3033	0.77	0.0625	0.18	0.48	0.77	0.7725
1.479	30.96	516	1.3519	0.78	0.065	0.19	0.515	0.7725	0.785
1.479	31.98	533	1.3688	0.775	0.065	0.1925	0.5225	0.765	0.775
1.479	33.0	550	1.3449	0.7725	0.065	0.2125	0.55	0.775	0.77
1.479	33.96	566	1.3758	0.77	0.0675	0.235	0.59	0.7625	0.77
1.479	34.98	583	1.4128	0.765	0.0675	0.2675	0.59	0.7725	0.7625
1.479	36.0	600	1.4091	0.7725	0.0675	0.31	0.5975	0.7825	0.775
1.479	36.96	616	1.4275	0.77	0.0675	0.295	0.6175	0.7725	0.7725
1.479	37.98	633	1.4912	0.7625	0.0675	0.3125	0.64	0.775	0.7625
1.479	39.0	650	1.4817	0.7725	0.065	0.3275	0.6425	0.77	0.7675
1.479	39.96	666	1.5270	0.7625	0.065	0.325	0.6575	0.775	0.765
1.479	40.98	683	1.5566	0.76	0.065	0.335	0.66	0.7675	0.7625
1.479	42.0	700	1.5411	0.76	0.065	0.36	0.6675	0.7625	0.7625
1.479	42.96	716	1.5643	0.755	0.065	0.37	0.6775	0.755	0.755
1.479	43.98	733	1.5261	0.765	0.065	0.365	0.6675	0.7575	0.765
1.479	45.0	750	1.5265	0.7625	0.065	0.365	0.67	0.765	0.765
1.479	45.96	766	1.5467	0.765	0.065	0.375	0.675	0.7725	0.765
1.479	46.98	783	1.5356	0.7725	0.065	0.3775	0.66	0.775	0.7675
1.479	48.0	800	1.5498	0.77	0.07	0.3825	0.6725	0.77	0.77
1.479	48.96	816	1.5103	0.7775	0.065	0.38	0.69	0.7725	0.775
1.479	49.98	833	1.5348	0.78	0.065	0.38	0.6825	0.775	0.78
1.479	51.0	850	1.5598	0.7775	0.07	0.3875	0.675	0.7675	0.775
1.479	51.96	866	1.5589	0.7775	0.065	0.4	0.6875	0.7675	0.7775
1.479	52.98	883	1.5564	0.775	0.065	0.405	0.69	0.775	0.775
1.479	54.0	900	1.5676	0.7725	0.065	0.3975	0.6875	0.7725	0.775
1.479	54.96	916	1.5690	0.7725	0.065	0.4025	0.6925	0.775	0.7725
1.479	55.98	933	1.5814	0.7725	0.065	0.405	0.695	0.7675	0.77
1.479	57.0	950	1.5753	0.775	0.0675	0.405	0.695	0.7725	0.775
1.479	57.6	960	1.5748	0.775	0.0675	0.405	0.6925	0.7725	0.775

Framework versions

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