2024-01-09_one_stage_subgraphs_weighted_txt_vision_enc_all_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: 1.1363
• Accuracy: 0.78
• Exit 0 Accuracy: 0.1575
• Exit 1 Accuracy: 0.2375
• Exit 2 Accuracy: 0.3475
• Exit 3 Accuracy: 0.27
• Exit 4 Accuracy: 0.3275
• Exit 5 Accuracy: 0.51
• Exit 6 Accuracy: 0.6075
• Exit 7 Accuracy: 0.7225
• Exit 8 Accuracy: 0.7575
• Exit 9 Accuracy: 0.76
• Exit 10 Accuracy: 0.735
• Exit 11 Accuracy: 0.7625
• Exit 12 Accuracy: 0.785
• Exit 13 Accuracy: 0.7775

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	Exit 5 Accuracy	Exit 6 Accuracy	Exit 7 Accuracy	Exit 8 Accuracy	Exit 9 Accuracy	Exit 10 Accuracy	Exit 11 Accuracy	Exit 12 Accuracy	Exit 13 Accuracy
No log	0.96	16	2.6689	0.185	0.0625	0.09	0.06	0.085	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.1075	0.0725
No log	1.98	33	2.4891	0.2725	0.0875	0.1175	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.0625	0.065	0.15
No log	3.0	50	2.2773	0.3425	0.1075	0.125	0.09	0.1125	0.0625	0.0625	0.0625	0.065	0.0625	0.0625	0.0625	0.0625	0.0925	0.2375
No log	3.96	66	2.0666	0.4	0.1	0.1175	0.0925	0.115	0.0625	0.0775	0.0625	0.065	0.0625	0.0625	0.0625	0.0625	0.065	0.305
No log	4.98	83	1.8597	0.485	0.12	0.12	0.0775	0.1225	0.0625	0.085	0.065	0.1125	0.0625	0.0625	0.0625	0.0625	0.0975	0.3275
No log	6.0	100	1.5470	0.5875	0.1225	0.135	0.1075	0.12	0.0625	0.0625	0.095	0.18	0.0825	0.0625	0.0625	0.0625	0.125	0.4425
No log	6.96	116	1.3384	0.6575	0.13	0.145	0.115	0.0825	0.065	0.0725	0.0925	0.19	0.1725	0.0625	0.0625	0.0625	0.12	0.5825
No log	7.98	133	1.1479	0.7125	0.1225	0.1475	0.11	0.135	0.0725	0.0925	0.12	0.2425	0.265	0.0625	0.0625	0.0625	0.12	0.63
No log	9.0	150	1.0184	0.745	0.1175	0.16	0.1575	0.1425	0.0775	0.095	0.1525	0.315	0.3775	0.0625	0.0625	0.0625	0.1225	0.68
No log	9.96	166	0.9324	0.7725	0.1225	0.1625	0.1425	0.13	0.095	0.1225	0.18	0.31	0.465	0.0625	0.0625	0.0625	0.1225	0.7225
No log	10.98	183	0.9027	0.755	0.125	0.165	0.1775	0.135	0.0975	0.125	0.24	0.405	0.515	0.0625	0.0625	0.0625	0.145	0.7325
No log	12.0	200	0.8841	0.7725	0.1225	0.1725	0.2	0.1425	0.095	0.115	0.3075	0.4525	0.5375	0.0625	0.0625	0.0625	0.2325	0.7525
No log	12.96	216	0.8951	0.7625	0.125	0.1725	0.2325	0.0625	0.115	0.12	0.3675	0.505	0.5425	0.0625	0.0625	0.085	0.2975	0.76
No log	13.98	233	0.9242	0.75	0.135	0.1825	0.1825	0.12	0.1675	0.13	0.3325	0.485	0.6225	0.0625	0.0625	0.0975	0.425	0.75
No log	15.0	250	0.8407	0.7775	0.13	0.185	0.165	0.11	0.1075	0.1275	0.3575	0.51	0.62	0.0625	0.0625	0.0975	0.455	0.7875
No log	15.96	266	1.0176	0.7325	0.165	0.19	0.1475	0.0925	0.2	0.1075	0.445	0.4625	0.6225	0.0625	0.0625	0.1375	0.5	0.7375
No log	16.98	283	0.9081	0.7825	0.15	0.19	0.17	0.095	0.21	0.125	0.455	0.5225	0.645	0.135	0.0925	0.2125	0.6275	0.79
No log	18.0	300	0.9397	0.7775	0.1425	0.1975	0.235	0.1125	0.195	0.1775	0.4225	0.56	0.665	0.1175	0.0625	0.195	0.665	0.785
No log	18.96	316	1.0205	0.765	0.1375	0.205	0.155	0.1	0.245	0.2225	0.4825	0.6275	0.6925	0.2175	0.37	0.17	0.71	0.7675
No log	19.98	333	0.9132	0.7975	0.1425	0.2	0.2525	0.1225	0.2775	0.3125	0.5375	0.6275	0.695	0.2175	0.305	0.2775	0.755	0.7975
No log	21.0	350	0.9610	0.78	0.1525	0.1975	0.2	0.1675	0.265	0.3225	0.5275	0.62	0.69	0.3975	0.285	0.435	0.7425	0.78
No log	21.96	366	0.9429	0.785	0.155	0.1975	0.2325	0.1425	0.2675	0.36	0.4975	0.645	0.7175	0.54	0.3725	0.55	0.775	0.78
No log	22.98	383	0.9549	0.795	0.1525	0.205	0.2625	0.16	0.29	0.3375	0.535	0.6525	0.72	0.5025	0.3675	0.635	0.765	0.795
No log	24.0	400	0.9764	0.7775	0.155	0.205	0.235	0.19	0.275	0.39	0.515	0.655	0.725	0.625	0.4275	0.615	0.7725	0.7825
No log	24.96	416	0.9804	0.7775	0.16	0.2025	0.32	0.16	0.2725	0.4125	0.5225	0.6425	0.7325	0.6375	0.4425	0.685	0.7775	0.7825
No log	25.98	433	0.9886	0.775	0.165	0.2075	0.3025	0.17	0.355	0.4875	0.5325	0.665	0.7325	0.6925	0.52	0.6175	0.7775	0.7775
No log	27.0	450	0.9906	0.7775	0.145	0.21	0.28	0.2075	0.33	0.4225	0.52	0.6675	0.7375	0.705	0.59	0.6675	0.7725	0.775
No log	27.96	466	1.0077	0.7875	0.1475	0.21	0.2625	0.23	0.3525	0.445	0.515	0.665	0.74	0.695	0.5975	0.6625	0.78	0.785
No log	28.98	483	1.0239	0.7775	0.1625	0.2075	0.3325	0.2075	0.35	0.455	0.55	0.6725	0.7425	0.72	0.6275	0.6925	0.7725	0.7825
0.4097	30.0	500	1.0323	0.785	0.16	0.21	0.375	0.3	0.365	0.49	0.555	0.68	0.7425	0.73	0.5575	0.705	0.7825	0.785
0.4097	30.96	516	1.0327	0.785	0.1525	0.2175	0.2975	0.205	0.3425	0.4625	0.55	0.68	0.74	0.7275	0.6	0.6925	0.785	0.7875
0.4097	31.98	533	1.0470	0.775	0.1575	0.22	0.3275	0.2575	0.36	0.51	0.56	0.695	0.7375	0.7275	0.5925	0.6725	0.785	0.775
0.4097	33.0	550	1.0538	0.785	0.17	0.2225	0.335	0.2775	0.375	0.5075	0.5575	0.69	0.7425	0.73	0.5725	0.7125	0.785	0.78
0.4097	33.96	566	1.0645	0.7775	0.155	0.22	0.3225	0.2425	0.3425	0.4675	0.555	0.7025	0.745	0.725	0.56	0.7175	0.7775	0.78
0.4097	34.98	583	1.0745	0.7825	0.155	0.23	0.305	0.2625	0.3725	0.4575	0.5625	0.7025	0.7375	0.7325	0.6275	0.735	0.7775	0.7775
0.4097	36.0	600	1.1022	0.7775	0.155	0.2275	0.305	0.245	0.3675	0.4775	0.5725	0.7025	0.7375	0.735	0.6425	0.7425	0.775	0.78
0.4097	36.96	616	1.1185	0.7675	0.1575	0.23	0.3175	0.2825	0.3575	0.48	0.575	0.7	0.74	0.7325	0.6425	0.7475	0.775	0.7725
0.4097	37.98	633	1.1017	0.7775	0.1575	0.23	0.31	0.2575	0.3575	0.4875	0.58	0.6975	0.75	0.735	0.6475	0.76	0.7775	0.775
0.4097	39.0	650	1.1036	0.7775	0.16	0.2325	0.3225	0.265	0.3375	0.49	0.5775	0.695	0.7425	0.72	0.6825	0.7375	0.7825	0.7775
0.4097	39.96	666	1.1200	0.775	0.165	0.2325	0.345	0.255	0.3325	0.4875	0.5875	0.7	0.7375	0.7175	0.6875	0.7425	0.7725	0.7775
0.4097	40.98	683	1.1207	0.775	0.165	0.23	0.3375	0.2575	0.3125	0.4975	0.5875	0.705	0.75	0.7275	0.6925	0.7525	0.7775	0.775
0.4097	42.0	700	1.1443	0.7725	0.165	0.2375	0.3475	0.245	0.3175	0.4925	0.5875	0.705	0.7525	0.72	0.675	0.7475	0.7675	0.77
0.4097	42.96	716	1.1497	0.7675	0.165	0.235	0.3575	0.2475	0.3325	0.51	0.585	0.7	0.7575	0.7275	0.695	0.7625	0.77	0.76
0.4097	43.98	733	1.1605	0.7625	0.17	0.2375	0.355	0.2675	0.3375	0.505	0.5875	0.7075	0.75	0.73	0.6975	0.7525	0.77	0.7625
0.4097	45.0	750	1.1612	0.7675	0.165	0.2375	0.345	0.2775	0.3225	0.5	0.6	0.705	0.755	0.7425	0.7025	0.755	0.775	0.77
0.4097	45.96	766	1.1632	0.7725	0.16	0.24	0.34	0.2625	0.3175	0.5	0.6	0.705	0.7575	0.7425	0.7225	0.7675	0.77	0.7725
0.4097	46.98	783	1.1555	0.7725	0.16	0.24	0.34	0.2725	0.3175	0.505	0.6	0.71	0.7525	0.7525	0.7175	0.76	0.7775	0.7775
0.4097	48.0	800	1.1437	0.775	0.155	0.2375	0.3325	0.2725	0.3175	0.4925	0.6	0.715	0.7575	0.7475	0.715	0.765	0.7825	0.7825
0.4097	48.96	816	1.1423	0.7775	0.155	0.2375	0.3325	0.2625	0.3175	0.5	0.6	0.715	0.76	0.75	0.7225	0.76	0.7775	0.7775
0.4097	49.98	833	1.1504	0.775	0.16	0.24	0.345	0.2625	0.32	0.505	0.6025	0.7175	0.75	0.75	0.7275	0.765	0.775	0.775
0.4097	51.0	850	1.1660	0.77	0.1575	0.24	0.3475	0.27	0.3175	0.5	0.605	0.715	0.755	0.7475	0.7325	0.7625	0.775	0.77
0.4097	51.96	866	1.1543	0.7775	0.16	0.2375	0.3425	0.27	0.325	0.5025	0.6075	0.72	0.76	0.7475	0.735	0.76	0.7825	0.775
0.4097	52.98	883	1.1363	0.7825	0.155	0.2375	0.3425	0.2875	0.325	0.505	0.605	0.7175	0.755	0.755	0.73	0.7625	0.7825	0.78
0.4097	54.0	900	1.1357	0.78	0.1575	0.2375	0.3425	0.275	0.3275	0.51	0.6075	0.7225	0.7575	0.755	0.735	0.7625	0.785	0.7775
0.4097	54.96	916	1.1354	0.78	0.1575	0.2375	0.3425	0.2725	0.33	0.51	0.6075	0.725	0.755	0.755	0.7375	0.7625	0.7875	0.78
0.4097	55.98	933	1.1356	0.78	0.1575	0.2375	0.3425	0.2725	0.3275	0.51	0.6075	0.725	0.7575	0.76	0.7375	0.7625	0.7875	0.78
0.4097	57.0	950	1.1362	0.7825	0.1575	0.2375	0.3475	0.27	0.3275	0.51	0.6075	0.7225	0.7575	0.76	0.735	0.7625	0.785	0.7775
0.4097	57.6	960	1.1363	0.78	0.1575	0.2375	0.3475	0.27	0.3275	0.51	0.6075	0.7225	0.7575	0.76	0.735	0.7625	0.785	0.7775

Framework versions

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