Swin-DMAE-H-DA-REVAL-80

This model is a fine-tuned version of microsoft/swinv2-tiny-patch4-window8-256 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 1.0766
• Accuracy: 0.75

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 80

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.608	0.97	22	1.6091	0.25
1.5899	1.98	45	1.5960	0.1923
1.4759	2.99	68	1.4430	0.3462
1.1012	4.0	91	1.3213	0.5192
0.8965	4.97	113	1.1938	0.4231
0.7214	5.98	136	1.1870	0.4615
0.6757	6.99	159	1.2117	0.5
0.5529	8.0	182	1.1976	0.4615
0.5279	8.97	204	1.1250	0.5192
0.4701	9.98	227	1.0999	0.5577
0.3721	10.99	250	0.7842	0.6538
0.3631	12.0	273	1.1728	0.6154
0.3384	12.97	295	1.2413	0.5769
0.2531	13.98	318	0.9144	0.6346
0.2753	14.99	341	0.8959	0.6923
0.2611	16.0	364	1.1399	0.6538
0.2072	16.97	386	1.0732	0.7115
0.2532	17.98	409	1.1922	0.7115
0.1633	18.99	432	1.0600	0.6731
0.1946	20.0	455	1.2289	0.6538
0.2214	20.97	477	1.3591	0.6731
0.1666	21.98	500	1.0736	0.7115
0.141	22.99	523	1.0315	0.6923
0.1275	24.0	546	1.0766	0.75
0.136	24.97	568	1.1796	0.7115
0.1402	25.98	591	1.0339	0.7115
0.1336	26.99	614	1.3446	0.6154
0.1218	28.0	637	1.2967	0.7115
0.1034	28.97	659	1.5955	0.6538
0.1196	29.98	682	1.5721	0.5769
0.1368	30.99	705	1.8208	0.6346
0.1477	32.0	728	1.4237	0.6923
0.1299	32.97	750	1.4061	0.7115
0.1111	33.98	773	1.6664	0.6346
0.068	34.99	796	1.7432	0.6538
0.1142	36.0	819	1.4518	0.6923
0.1258	36.97	841	1.7217	0.6346
0.1055	37.98	864	1.6348	0.6154
0.1049	38.99	887	1.8378	0.6346
0.0822	40.0	910	1.6760	0.6731
0.1114	40.97	932	1.7310	0.6346
0.0704	41.98	955	1.7105	0.6538
0.0983	42.99	978	1.8320	0.5962
0.0909	44.0	1001	1.5632	0.6346
0.0991	44.97	1023	1.7606	0.6731
0.0658	45.98	1046	1.5927	0.6538
0.0412	46.99	1069	1.4660	0.6538
0.0919	48.0	1092	1.3294	0.6731
0.0726	48.97	1114	1.5551	0.6346
0.0554	49.98	1137	1.7157	0.6154
0.0585	50.99	1160	1.8280	0.5962
0.0607	52.0	1183	1.6142	0.6538
0.0719	52.97	1205	1.9924	0.5962
0.0877	53.98	1228	1.7806	0.6346
0.0743	54.99	1251	1.9820	0.6538
0.0464	56.0	1274	1.9449	0.6346
0.077	56.97	1296	1.6826	0.6923
0.073	57.98	1319	1.7594	0.6538
0.0623	58.99	1342	1.8303	0.6346
0.0383	60.0	1365	1.8124	0.6154
0.0526	60.97	1387	1.8164	0.6923
0.0679	61.98	1410	1.8586	0.6731
0.0625	62.99	1433	1.9150	0.6346
0.0482	64.0	1456	1.9622	0.6346
0.0646	64.97	1478	1.9476	0.6154
0.0594	65.98	1501	1.5958	0.6923
0.0568	66.99	1524	1.8275	0.6731
0.0662	68.0	1547	1.7576	0.6731
0.0428	68.97	1569	1.9325	0.6538
0.0433	69.98	1592	1.8206	0.6731
0.0511	70.99	1615	1.9029	0.6538
0.0502	72.0	1638	1.8821	0.6538
0.0544	72.97	1660	1.9535	0.6538
0.0399	73.98	1683	1.8455	0.6538
0.0561	74.99	1706	1.8290	0.6538
0.041	76.0	1729	1.8427	0.6538
0.0582	76.97	1751	1.8591	0.6538
0.0315	77.36	1760	1.8612	0.6538

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0