vit-base-patch16-224-FV2-finetuned-memes

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: 0.5458
Accuracy: 0.8648
Precision: 0.8651
Recall: 0.8648
F1: 0.8646

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
0.994	0.99	20	0.7937	0.7257	0.7148	0.7257	0.7025
0.509	1.99	40	0.4634	0.8346	0.8461	0.8346	0.8303
0.2698	2.99	60	0.3851	0.8594	0.8619	0.8594	0.8586
0.1381	3.99	80	0.4186	0.8624	0.8716	0.8624	0.8634
0.0899	4.99	100	0.4038	0.8586	0.8624	0.8586	0.8594
0.0708	5.99	120	0.4170	0.8563	0.8612	0.8563	0.8580
0.0629	6.99	140	0.4414	0.8594	0.8599	0.8594	0.8585
0.0554	7.99	160	0.4617	0.8539	0.8563	0.8539	0.8550
0.0582	8.99	180	0.4712	0.8648	0.8667	0.8648	0.8651
0.0582	9.99	200	0.4753	0.8632	0.8647	0.8632	0.8636
0.0535	10.99	220	0.4653	0.8694	0.8690	0.8694	0.8684
0.0516	11.99	240	0.4937	0.8679	0.8692	0.8679	0.8681
0.0478	12.99	260	0.5109	0.8725	0.8741	0.8725	0.8718
0.0484	13.99	280	0.5144	0.8640	0.8660	0.8640	0.8647
0.0472	14.99	300	0.5249	0.8679	0.8688	0.8679	0.8678
0.043	15.99	320	0.5324	0.8709	0.8711	0.8709	0.8704
0.0473	16.99	340	0.5352	0.8648	0.8660	0.8648	0.8647
0.0502	17.99	360	0.5389	0.8694	0.8692	0.8694	0.8687
0.0489	18.99	380	0.5564	0.8648	0.8666	0.8648	0.8651
0.04	19.99	400	0.5458	0.8648	0.8651	0.8648	0.8646

Framework versions

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

作者:

paul

数据集大小:

327.43 MB