SW2-TO-DA

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: 0.2207
• Accuracy: 0.9194

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: 0.00015
• 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: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.4955	0.97	14	1.5580	0.0806
1.3943	2.0	29	1.1316	0.6452
1.0056	2.97	43	0.6407	0.7419
0.7744	4.0	58	0.4265	0.8710
0.6022	4.97	72	0.4361	0.8548
0.5854	6.0	87	0.5508	0.8065
0.4581	6.97	101	0.3124	0.8548
0.386	8.0	116	0.3169	0.8548
0.347	8.97	130	0.2207	0.9194
0.3873	10.0	145	0.5969	0.8226
0.3508	10.97	159	0.3425	0.8871
0.274	12.0	174	0.3376	0.8710
0.2615	12.97	188	0.4913	0.8710
0.3118	14.0	203	0.4034	0.8871
0.2205	14.97	217	0.3167	0.8710
0.2325	16.0	232	0.3043	0.8871
0.1914	16.97	246	0.4256	0.8226
0.1997	18.0	261	0.3769	0.8548
0.1752	18.97	275	0.5875	0.8548
0.1685	20.0	290	0.4104	0.8871
0.1736	20.97	304	0.5481	0.8548
0.1901	22.0	319	0.3800	0.9032
0.1426	22.97	333	0.4425	0.8871
0.1251	24.0	348	0.3374	0.9032
0.1326	24.97	362	0.3627	0.8871
0.1271	26.0	377	0.4768	0.8710
0.1835	26.97	391	0.5604	0.8710
0.1378	28.0	406	0.4131	0.8871
0.1349	28.97	420	0.5103	0.8548
0.0999	30.0	435	0.3723	0.9194
0.1198	30.97	449	0.5361	0.8710
0.1195	32.0	464	0.4194	0.8871
0.0766	32.97	478	0.4133	0.8871
0.0862	34.0	493	0.4239	0.9032
0.1048	34.97	507	0.4120	0.9194
0.0902	36.0	522	0.4408	0.9032
0.088	36.97	536	0.4436	0.9032
0.089	38.0	551	0.4648	0.9032
0.1089	38.62	560	0.4650	0.8871

Framework versions

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