cvt-21-finetuned-brs2

This model is a fine-tuned version of microsoft/cvt-21 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.6947
• Accuracy: 0.6604
• F1: 0.6087
• Precision (ppv): 0.5385
• Recall (sensitivity): 0.7
• Specificity: 0.6364
• Npv: 0.7778
• Auc: 0.6682

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision (ppv)	Recall (sensitivity)	Specificity	Npv	Auc
0.8177	1.89	100	0.7113	0.5283	0.5098	0.4194	0.65	0.4545	0.6818	0.5523
0.736	3.77	200	0.7178	0.5283	0.3902	0.3810	0.4	0.6061	0.625	0.5030
0.5978	5.66	300	0.6889	0.6038	0.5532	0.4815	0.65	0.5758	0.7308	0.6129
0.5576	7.55	400	0.7349	0.4717	0.5484	0.4048	0.85	0.2424	0.7273	0.5462
0.5219	9.43	500	0.6522	0.6038	0.4	0.4667	0.35	0.7576	0.6579	0.5538
0.5326	11.32	600	0.6665	0.6226	0.5238	0.5	0.55	0.6667	0.7097	0.6083
0.4381	13.21	700	0.7685	0.4717	0.5333	0.4	0.8	0.2727	0.6923	0.5364
0.5598	15.09	800	0.7212	0.5283	0.1935	0.2727	0.15	0.7576	0.5952	0.4538
0.6887	16.98	900	0.6985	0.6604	0.64	0.5333	0.8	0.5758	0.8261	0.6879
0.7594	18.87	1000	0.7040	0.5472	0.4286	0.4091	0.45	0.6061	0.6452	0.5280
0.2177	20.75	1100	0.8056	0.4528	0.5397	0.3953	0.85	0.2121	0.7	0.5311
0.4893	22.64	1200	0.8821	0.3396	0.3860	0.2973	0.55	0.2121	0.4375	0.3811
0.5994	24.53	1300	0.8059	0.5660	0.5660	0.4545	0.75	0.4545	0.75	0.6023
0.5179	26.42	1400	0.6750	0.6038	0.4615	0.4737	0.45	0.6970	0.6765	0.5735
0.198	28.3	1500	0.7448	0.3962	0.3333	0.2857	0.4	0.3939	0.52	0.3970
0.6536	30.19	1600	0.7555	0.5094	0.4583	0.3929	0.55	0.4848	0.64	0.5174
0.7558	32.08	1700	0.6664	0.5849	0.4762	0.4545	0.5	0.6364	0.6774	0.5682
0.4915	33.96	1800	0.9213	0.3962	0.5152	0.3696	0.85	0.1212	0.5714	0.4856
0.3661	35.85	1900	0.9202	0.4528	0.4912	0.3784	0.7	0.3030	0.625	0.5015
0.4838	37.74	2000	0.9297	0.4528	0.5085	0.3846	0.75	0.2727	0.6429	0.5114
0.8461	39.62	2100	0.9464	0.4717	0.5758	0.4130	0.95	0.1818	0.8571	0.5659
0.6937	41.51	2200	0.7129	0.5094	0.48	0.4	0.6	0.4545	0.6522	0.5273
0.6302	43.4	2300	0.6866	0.5849	0.6071	0.4722	0.85	0.4242	0.8235	0.6371
0.0793	45.28	2400	0.7791	0.5094	0.5517	0.4211	0.8	0.3333	0.7333	0.5667
0.464	47.17	2500	0.8116	0.4340	0.4444	0.3529	0.6	0.3333	0.5789	0.4667
0.6131	49.06	2600	0.5970	0.6226	0.5455	0.5	0.6	0.6364	0.7241	0.6182
0.6937	50.94	2700	0.8201	0.4340	0.4	0.3333	0.5	0.3939	0.5652	0.4470
0.6552	52.83	2800	0.7168	0.5660	0.5306	0.4483	0.65	0.5152	0.7083	0.5826
0.7749	54.72	2900	0.6875	0.5849	0.5217	0.4615	0.6	0.5758	0.7037	0.5879
0.9482	56.6	3000	0.6392	0.6226	0.6296	0.5	0.85	0.4848	0.8421	0.6674
0.2467	58.49	3100	0.6281	0.6038	0.5333	0.48	0.6	0.6061	0.7143	0.6030
0.2903	60.38	3200	0.7383	0.5472	0.5556	0.4412	0.75	0.4242	0.7368	0.5871
0.5859	62.26	3300	0.7191	0.6226	0.5652	0.5	0.65	0.6061	0.7407	0.6280
0.3815	64.15	3400	0.7469	0.5283	0.4444	0.4	0.5	0.5455	0.6429	0.5227
0.531	66.04	3500	0.7566	0.6226	0.5652	0.5	0.65	0.6061	0.7407	0.6280
0.3892	67.92	3600	0.8168	0.5660	0.5490	0.4516	0.7	0.4848	0.7273	0.5924
0.6487	69.81	3700	0.9077	0.4340	0.4643	0.3611	0.65	0.3030	0.5882	0.4765
0.5525	71.7	3800	0.6961	0.6038	0.5116	0.4783	0.55	0.6364	0.7	0.5932
0.3137	73.58	3900	1.0817	0.3774	0.4590	0.3415	0.7	0.1818	0.5	0.4409
0.3526	75.47	4000	0.7684	0.5472	0.5862	0.4474	0.85	0.3636	0.8	0.6068
0.5938	77.36	4100	0.8786	0.4340	0.4828	0.3684	0.7	0.2727	0.6	0.4864
0.2431	79.25	4200	0.8925	0.4151	0.4746	0.3590	0.7	0.2424	0.5714	0.4712
0.1021	81.13	4300	1.0740	0.4528	0.4727	0.3714	0.65	0.3333	0.6111	0.4917
0.3429	83.02	4400	0.7723	0.4906	0.5091	0.4	0.7	0.3636	0.6667	0.5318
0.3836	84.91	4500	0.7247	0.5472	0.5556	0.4412	0.75	0.4242	0.7368	0.5871
0.4099	86.79	4600	0.8508	0.4340	0.4828	0.3684	0.7	0.2727	0.6	0.4864
0.8264	88.68	4700	0.7682	0.5849	0.5769	0.4688	0.75	0.4848	0.7619	0.6174
0.1928	90.57	4800	0.8738	0.4906	0.5574	0.4146	0.85	0.2727	0.75	0.5614
0.3422	92.45	4900	0.8810	0.5660	0.5965	0.4595	0.85	0.3939	0.8125	0.6220
0.5524	94.34	5000	1.0801	0.3774	0.4923	0.3556	0.8	0.1212	0.5	0.4606
0.464	96.23	5100	0.9417	0.5283	0.5902	0.4390	0.9	0.3030	0.8333	0.6015
0.7182	98.11	5200	1.0335	0.4151	0.4746	0.3590	0.7	0.2424	0.5714	0.4712
0.604	100.0	5300	0.6947	0.6604	0.6087	0.5385	0.7	0.6364	0.7778	0.6682

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1+cu113
• Datasets 2.6.1
• Tokenizers 0.13.1