dino-large-2023_12_08-with_custom_head-imgsize1036

This model is a fine-tuned version of facebook/dinov2-large on the multilabel_complete_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.1017
• F1 Micro: 0.8547
• F1 Macro: 0.8214
• Roc Auc: 0.9081
• Accuracy: 0.5649
• Learning Rate: 0.0001

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: 0.01
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Roc Auc	Accuracy	Rate
0.2502	1.0	536	0.1990	0.6427	0.5121	0.7510	0.4030	0.01
0.2164	2.0	1072	1.1451	0.6769	0.6296	0.7895	0.4048	0.01
0.2149	3.0	1608	0.2339	0.6516	0.5251	0.7524	0.3994	0.01
0.2159	4.0	2144	0.1699	0.7376	0.6028	0.8297	0.4116	0.01
0.2171	5.0	2680	0.1650	0.7364	0.6304	0.8166	0.4394	0.01
0.2166	6.0	3216	0.1748	0.6853	0.5123	0.7773	0.3923	0.01
0.2081	7.0	3752	0.1636	0.7455	0.6129	0.8291	0.4348	0.01
0.2111	8.0	4288	0.1651	0.7543	0.6445	0.8476	0.4277	0.01
0.209	9.0	4824	0.1750	0.7062	0.6522	0.7966	0.4359	0.01
0.2107	10.0	5360	0.1751	0.7244	0.5924	0.8146	0.3830	0.01
0.2162	11.0	5896	0.2229	0.7506	0.6780	0.8475	0.4252	0.01
0.2153	12.0	6432	0.1740	0.7501	0.6543	0.8550	0.4105	0.01
0.2197	13.0	6968	0.1745	0.7487	0.6605	0.8572	0.4187	0.01
0.18	14.0	7504	0.1348	0.8036	0.7455	0.8731	0.5059	0.001
0.164	15.0	8040	0.1308	0.8160	0.7783	0.8844	0.5173	0.001
0.162	16.0	8576	0.1305	0.8188	0.7530	0.8764	0.5202	0.001
0.1548	17.0	9112	0.1242	0.8291	0.7887	0.8945	0.5248	0.001
0.1532	18.0	9648	0.1247	0.8292	0.7823	0.8934	0.5227	0.001
0.152	19.0	10184	0.1272	0.8238	0.7688	0.8832	0.5280	0.001
0.1479	20.0	10720	0.1239	0.8280	0.7783	0.8834	0.5288	0.001
0.1483	21.0	11256	0.1376	0.8361	0.7914	0.8919	0.5341	0.001
0.1448	22.0	11792	0.1267	0.8292	0.7774	0.8842	0.5380	0.001
0.1456	23.0	12328	0.1217	0.8334	0.7914	0.8883	0.5448	0.001
0.1441	24.0	12864	0.1193	0.8283	0.7852	0.8801	0.5380	0.001
0.1406	25.0	13400	0.1185	0.8392	0.8020	0.8988	0.5341	0.001
0.1416	26.0	13936	0.1295	0.8351	0.7851	0.8889	0.5441	0.001
0.1417	27.0	14472	0.1390	0.8287	0.7699	0.8808	0.5305	0.001
0.142	28.0	15008	0.1256	0.8328	0.7857	0.8888	0.5441	0.001
0.14	29.0	15544	0.1268	0.8291	0.7759	0.8815	0.5359	0.001
0.1415	30.0	16080	0.1374	0.8240	0.7675	0.8722	0.5420	0.001
0.1414	31.0	16616	0.1281	0.8310	0.7795	0.8838	0.5406	0.001
0.1349	32.0	17152	0.1144	0.8389	0.7927	0.8892	0.5513	0.0001
0.1294	33.0	17688	0.1097	0.8414	0.7991	0.8915	0.5534	0.0001
0.1281	34.0	18224	0.1160	0.8425	0.7982	0.8925	0.5520	0.0001
0.1274	35.0	18760	0.1244	0.8441	0.7999	0.8935	0.5577	0.0001
0.1243	36.0	19296	0.1100	0.8434	0.7991	0.8898	0.5559	0.0001
0.1231	37.0	19832	0.1073	0.8485	0.8086	0.8989	0.5641	0.0001
0.1245	38.0	20368	0.1092	0.8456	0.8054	0.8916	0.5602	0.0001
0.1197	39.0	20904	0.1069	0.8483	0.8112	0.9002	0.5623	0.0001
0.1242	40.0	21440	0.1065	0.8468	0.8081	0.8949	0.5638	0.0001
0.1167	41.0	21976	0.1083	0.8462	0.8043	0.8934	0.5591	0.0001
0.1179	42.0	22512	0.1072	0.8505	0.8090	0.8978	0.5691	0.0001
0.1186	43.0	23048	0.1179	0.8483	0.8072	0.8945	0.5613	0.0001
0.1174	44.0	23584	0.1068	0.8477	0.8080	0.8946	0.5656	0.0001
0.1153	45.0	24120	0.1047	0.8534	0.8193	0.9025	0.5716	0.0001
0.1167	46.0	24656	0.1062	0.8535	0.8229	0.9080	0.5666	0.0001
0.1162	47.0	25192	0.1060	0.8522	0.8180	0.9006	0.5695	0.0001
0.1145	48.0	25728	0.1041	0.8529	0.8154	0.9002	0.5745	0.0001
0.1143	49.0	26264	0.1033	0.8542	0.8182	0.9043	0.5652	0.0001
0.1129	50.0	26800	0.1054	0.8508	0.8102	0.8956	0.5720	0.0001

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1