20E-affecthq

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

• Loss: 0.8271
• Accuracy: 0.7188
• Precision: 0.7220
• Recall: 0.7188
• F1: 0.7190

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-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 17
• 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

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
1.9149	1.0	194	1.8887	0.3750	0.3413	0.3750	0.3045
1.2903	2.0	388	1.2485	0.5792	0.5726	0.5792	0.5526
1.071	3.0	582	1.0587	0.6321	0.6258	0.6321	0.6228
1.0185	4.0	776	0.9817	0.6617	0.6584	0.6617	0.6553
0.894	5.0	970	0.9293	0.6869	0.6872	0.6869	0.6820
0.8283	6.0	1164	0.8881	0.6936	0.6929	0.6936	0.6905
0.8185	7.0	1358	0.8659	0.6982	0.7011	0.6982	0.6988
0.7499	8.0	1552	0.8558	0.7046	0.7050	0.7046	0.7021
0.7219	9.0	1746	0.8399	0.7124	0.7165	0.7124	0.7127
0.7382	10.0	1940	0.8300	0.7159	0.7184	0.7159	0.7145
0.6392	11.0	2134	0.8329	0.7088	0.7135	0.7088	0.7095
0.6549	12.0	2328	0.8297	0.7133	0.7135	0.7133	0.7120
0.6762	13.0	2522	0.8180	0.7156	0.7162	0.7156	0.7153
0.5937	14.0	2716	0.8271	0.7188	0.7220	0.7188	0.7190
0.569	15.0	2910	0.8245	0.7178	0.7175	0.7178	0.7165
0.5623	16.0	3104	0.8228	0.7165	0.7153	0.7165	0.7157
0.5291	17.0	3298	0.8238	0.7162	0.7165	0.7162	0.7156
0.5775	18.0	3492	0.8246	0.7153	0.7162	0.7153	0.7151
0.545	19.0	3686	0.8257	0.7178	0.7192	0.7178	0.7174
0.5409	20.0	3880	0.8245	0.7178	0.7187	0.7178	0.7177

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2