SWv2-DMAE-H-4-rp-clean-fix-U-40-Cross-1

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: 0.6198
• Accuracy: 0.8452

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: 4e-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: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.6089	0.98	12	1.6063	0.2024
1.6028	1.96	24	1.5805	0.2024
1.5642	2.94	36	1.5728	0.2024
1.5232	4.0	49	1.5104	0.2024
1.4013	4.98	61	1.3106	0.5119
1.2544	5.96	73	1.0733	0.6190
1.1223	6.94	85	0.8374	0.7381
0.977	8.0	98	0.7359	0.75
0.8368	8.98	110	0.7201	0.7738
0.7628	9.96	122	0.6245	0.7857
0.7063	10.94	134	0.6683	0.8095
0.6462	12.0	147	0.6519	0.7619
0.6588	12.98	159	0.5841	0.8095
0.5777	13.96	171	0.6083	0.7976
0.5363	14.94	183	0.6363	0.7738
0.5545	16.0	196	0.7957	0.7143
0.5073	16.98	208	0.5762	0.8095
0.4373	17.96	220	0.5577	0.7976
0.3747	18.94	232	0.5768	0.7738
0.4577	20.0	245	0.5946	0.8095
0.437	20.98	257	0.6457	0.7738
0.3996	21.96	269	0.6653	0.7976
0.336	22.94	281	0.6262	0.8095
0.3371	24.0	294	0.6686	0.8095
0.3571	24.98	306	0.6868	0.7976
0.3465	25.96	318	0.6071	0.8214
0.3324	26.94	330	0.6163	0.8333
0.3161	28.0	343	0.6440	0.8333
0.2851	28.98	355	0.6198	0.8452
0.304	29.96	367	0.6296	0.8452
0.3175	30.94	379	0.6400	0.8452
0.2899	32.0	392	0.6439	0.8333
0.2674	32.98	404	0.6620	0.8452
0.264	33.96	416	0.6708	0.8452
0.2727	34.94	428	0.6524	0.8333
0.2818	36.0	441	0.6483	0.8333
0.2596	36.98	453	0.6505	0.8452
0.2498	37.96	465	0.6515	0.8333
0.2181	38.94	477	0.6522	0.8333
0.2428	39.18	480	0.6523	0.8333

Framework versions

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