face_predict

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: 1.2322
• Accuracy: 0.5625

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 6
• total_train_batch_size: 192
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.9	3	2.0747	0.1187
No log	1.8	6	2.0728	0.1375
2.0713	3.0	10	2.0449	0.2
2.0713	3.9	13	2.0225	0.2562
2.0713	4.8	16	1.9779	0.2938
1.9642	6.0	20	1.8985	0.3688
1.9642	6.9	23	1.8440	0.4188
1.9642	7.8	26	1.7593	0.4437
1.7442	9.0	30	1.6551	0.4875
1.7442	9.9	33	1.5996	0.4875
1.7442	10.8	36	1.5324	0.5188
1.5402	12.0	40	1.5053	0.525
1.5402	12.9	43	1.4543	0.5188
1.5402	13.8	46	1.4335	0.5188
1.4064	15.0	50	1.3768	0.5938
1.4064	15.9	53	1.3583	0.6
1.4064	16.8	56	1.3464	0.575
1.2844	18.0	60	1.3245	0.6125
1.2844	18.9	63	1.3265	0.5563
1.2844	19.8	66	1.2899	0.5813
1.1834	21.0	70	1.2863	0.5625
1.1834	21.9	73	1.2939	0.5687
1.1834	22.8	76	1.2508	0.5938
1.1046	24.0	80	1.2604	0.5563
1.1046	24.9	83	1.2344	0.6062
1.1046	25.8	86	1.2124	0.6125
1.0379	27.0	90	1.2053	0.6312
1.0379	27.9	93	1.3067	0.5375
1.0379	28.8	96	1.2247	0.5875
1.0064	30.0	100	1.2060	0.625
1.0064	30.9	103	1.2308	0.575
1.0064	31.8	106	1.1936	0.6188
0.9611	33.0	110	1.2257	0.5938
0.9611	33.9	113	1.2302	0.5563
0.9611	34.8	116	1.2172	0.6
0.9351	36.0	120	1.2355	0.55

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1