dino-vitb16-finetuned-galaxy10-decals

This model is a fine-tuned version of facebook/dino-vitb16 on the matthieulel/galaxy10_decals dataset. It achieves the following results on the evaluation set:

• Loss: 1.3956
• Accuracy: 0.5062

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: 1e-07
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.8951	0.9940	124	2.7751	0.0868
2.5724	1.9960	249	2.5352	0.0817
2.2714	2.9980	374	2.2738	0.1285
2.0997	4.0	499	2.0739	0.2182
1.9182	4.9940	623	1.9210	0.3027
1.8259	5.9960	748	1.8074	0.3608
1.7096	6.9980	873	1.7133	0.3923
1.667	8.0	998	1.6444	0.4177
1.5932	8.9940	1122	1.5951	0.4335
1.5421	9.9960	1247	1.5479	0.4464
1.5117	10.9980	1372	1.5130	0.4600
1.489	12.0	1497	1.4821	0.4735
1.4881	12.9940	1621	1.4592	0.4786
1.4648	13.9960	1746	1.4420	0.4865
1.4387	14.9980	1871	1.4277	0.4927
1.46	16.0	1996	1.4154	0.4994
1.4393	16.9940	2120	1.4056	0.5
1.4471	17.9960	2245	1.3995	0.5045
1.4085	18.9980	2370	1.3956	0.5062
1.4227	19.8798	2480	1.3948	0.5056

Framework versions

• Transformers 4.40.1
• Pytorch 1.12.1+cu116
• Datasets 2.19.0
• Tokenizers 0.19.1