vit-base-patch16-224-AHB-against-NotAHB

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

• Loss: 0.1588
• Accuracy: 0.9538

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: 0.0005
• 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: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.8421	4	0.7464	0.6462
No log	1.8947	9	0.4871	0.8154
0.7361	2.9474	14	0.6255	0.7385
0.7361	4.0	19	0.3812	0.8308
0.68	4.8421	23	0.3080	0.8615
0.68	5.8947	28	0.2427	0.9231
0.4598	6.9474	33	0.2114	0.9077
0.4598	8.0	38	0.2383	0.9231
0.4265	8.8421	42	0.3264	0.8462
0.4265	9.8947	47	0.2211	0.8769
0.3265	10.9474	52	0.2056	0.9077
0.3265	12.0	57	0.4595	0.7538
0.3282	12.8421	61	0.1888	0.9385
0.3282	13.8947	66	0.1588	0.9538
0.296	14.9474	71	0.3073	0.8769
0.296	16.0	76	0.1888	0.9231
0.2679	16.8421	80	0.2230	0.9077

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1