vit-molecul

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

• Loss: 0.5737
• Accuracy: 0.71
• F1: 0.7086

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: 3e-06
• train_batch_size: 50
• eval_batch_size: 50
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.723	1.0	8	0.6790	0.61	0.6096
0.6915	2.0	16	0.6661	0.62	0.5924
0.6689	3.0	24	0.6470	0.69	0.6892
0.6517	4.0	32	0.6356	0.64	0.6377
0.6368	5.0	40	0.6289	0.72	0.7199
0.621	6.0	48	0.6217	0.73	0.7293
0.6061	7.0	56	0.6197	0.69	0.6862
0.5924	8.0	64	0.6087	0.73	0.7293
0.5767	9.0	72	0.6003	0.72	0.7199
0.5633	10.0	80	0.5953	0.72	0.7196
0.5491	11.0	88	0.5885	0.72	0.7199
0.5351	12.0	96	0.5869	0.71	0.7100
0.5239	13.0	104	0.5867	0.7	0.6995
0.5118	14.0	112	0.5804	0.71	0.7100
0.502	15.0	120	0.5752	0.71	0.7100
0.4942	16.0	128	0.5738	0.72	0.7199
0.4885	17.0	136	0.5771	0.71	0.7086
0.4831	18.0	144	0.5751	0.71	0.7086
0.4793	19.0	152	0.5743	0.71	0.7086
0.4774	20.0	160	0.5737	0.71	0.7086

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.1
• Tokenizers 0.13.3