galaxy10_decals_vit

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5782
• Accuracy: 0.8391

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.216	1.0	363	1.1126	0.7115
0.8912	2.0	727	0.8005	0.7688
0.7469	3.0	1091	0.6824	0.7920
0.7641	4.0	1455	0.6010	0.8128
0.6819	5.0	1818	0.5770	0.8190
0.5211	6.0	2182	0.5590	0.8167
0.6417	7.0	2546	0.5780	0.8113
0.5331	8.0	2910	0.5579	0.8268
0.4962	9.0	3273	0.5375	0.8415
0.4047	10.0	3637	0.5290	0.8337
0.3155	11.0	4001	0.5136	0.8430
0.3577	12.0	4365	0.5641	0.8283
0.3345	13.0	4728	0.5634	0.8306
0.299	14.0	5092	0.5855	0.8260
0.2879	15.0	5456	0.5459	0.8407
0.2822	16.0	5820	0.5681	0.8368
0.3435	17.0	6183	0.5806	0.8322
0.2899	18.0	6547	0.5857	0.8368
0.2323	19.0	6911	0.5781	0.8337
0.2266	19.96	7260	0.5782	0.8391

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2