google-vit-base-patch16-224-cartoon-face-recognition

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: 0.3707
• Accuracy: 0.9005
• Precision: 0.9066
• Recall: 0.9005
• F1: 0.8984

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: 0.00012
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• 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	Precision	Recall	F1
No log	0.89	6	0.5459	0.8611	0.8683	0.8611	0.8577
0.0812	1.89	12	0.4703	0.8796	0.8833	0.8796	0.8764
0.0812	2.89	18	0.4430	0.8935	0.8969	0.8935	0.8906
0.0307	3.89	24	0.4045	0.8819	0.8849	0.8819	0.8767
0.0091	4.89	30	0.3672	0.9005	0.9025	0.9005	0.8980
0.0091	5.89	36	0.3841	0.9028	0.9125	0.9028	0.9011
0.0043	6.89	42	0.3926	0.9005	0.9073	0.9005	0.8972
0.0043	7.89	48	0.3786	0.8958	0.9005	0.8958	0.8931
0.0031	8.89	54	0.3791	0.9028	0.9091	0.9028	0.9007
0.002	9.89	60	0.3677	0.9028	0.9106	0.9028	0.9001
0.002	10.89	66	0.3740	0.9028	0.9099	0.9028	0.9007
0.0027	11.89	72	0.3869	0.8981	0.9043	0.8981	0.8956
0.0027	12.89	78	0.3801	0.8981	0.9021	0.8981	0.8954
0.004	13.89	84	0.3674	0.9051	0.9113	0.9051	0.9028
0.0024	14.89	90	0.3620	0.9051	0.9096	0.9051	0.9027
0.0024	15.89	96	0.3670	0.9028	0.9089	0.9028	0.9006
0.0021	16.89	102	0.3827	0.9005	0.9065	0.9005	0.8980
0.0021	17.89	108	0.3748	0.8981	0.9049	0.8981	0.8958
0.0022	18.89	114	0.3825	0.9028	0.9101	0.9028	0.9006
0.0019	19.89	120	0.3707	0.9005	0.9066	0.9005	0.8984

Framework versions

• Transformers 4.24.0.dev0
• Pytorch 1.11.0+cu102
• Datasets 2.6.1
• Tokenizers 0.13.1