swin-camdeboo

This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on the camdeboo dataset. It achieves the following results on the evaluation set:

• Loss: 1.7505
• Accuracy: 0.5693

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: 0.0002
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.6714	0.0647	100	2.1798	0.3780
1.3024	0.1294	200	2.0776	0.4628
1.2379	0.1942	300	2.1343	0.4699
0.921	0.2589	400	2.0805	0.4909
1.2239	0.3236	500	1.8752	0.5048
1.1477	0.3883	600	2.1270	0.4671
1.0999	0.4531	700	1.8343	0.5404
0.8614	0.5178	800	2.1993	0.4814
1.4409	0.5825	900	2.0452	0.5436
1.1298	0.6472	1000	2.0431	0.5234
0.6623	0.7120	1100	1.8523	0.5626
0.5556	0.7767	1200	1.9374	0.5745
0.6044	0.8414	1300	2.1097	0.5293
0.7965	0.9061	1400	1.7505	0.5693
0.6878	0.9709	1500	2.0018	0.5590
0.763	1.0356	1600	2.1517	0.5527
0.3674	1.1003	1700	2.5025	0.5642
0.6865	1.1650	1800	2.0337	0.5654
0.3911	1.2298	1900	2.1313	0.5852
0.6468	1.2945	2000	2.4708	0.5689
0.4501	1.3592	2100	2.2926	0.5911
0.2603	1.4239	2200	2.3958	0.6026
0.327	1.4887	2300	2.0308	0.5963
0.5728	1.5534	2400	2.5120	0.5860
0.3978	1.6181	2500	2.2434	0.6086
0.3691	1.6828	2600	2.2805	0.6208
0.443	1.7476	2700	2.2768	0.6097
0.5276	1.8123	2800	2.2888	0.6193
0.3785	1.8770	2900	2.3886	0.5983
0.5051	1.9417	3000	2.3341	0.5975

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1