all-observation-type

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

• Loss: 0.0077
• F1: 0.0913

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1
0.0726	1.1628	100	0.0660	0.0
0.0264	2.3256	200	0.0247	0.0
0.0161	3.4884	300	0.0165	0.0
0.0133	4.6512	400	0.0135	0.0
0.0124	5.8140	500	0.0120	0.0
0.011	6.9767	600	0.0112	0.0
0.0114	8.1395	700	0.0107	0.0
0.0109	9.3023	800	0.0103	0.0
0.0096	10.4651	900	0.0102	0.0
0.0099	11.6279	1000	0.0098	0.0
0.0089	12.7907	1100	0.0094	0.0
0.0091	13.9535	1200	0.0093	0.0
0.0081	15.1163	1300	0.0089	0.0
0.0073	16.2791	1400	0.0089	0.0
0.0071	17.4419	1500	0.0085	0.0
0.0068	18.6047	1600	0.0082	0.0183
0.0064	19.7674	1700	0.0082	0.0365
0.0061	20.9302	1800	0.0086	0.0091
0.0054	22.0930	1900	0.0082	0.0594
0.0051	23.2558	2000	0.0080	0.0502
0.0048	24.4186	2100	0.0079	0.0639
0.0045	25.5814	2200	0.0080	0.0639
0.0036	26.7442	2300	0.0079	0.1027
0.0038	27.9070	2400	0.0079	0.1027
0.0032	29.0698	2500	0.0077	0.0913
0.004	30.2326	2600	0.0079	0.1027
0.003	31.3953	2700	0.0081	0.0936
0.0029	32.5581	2800	0.0080	0.0890
0.0033	33.7209	2900	0.0081	0.0845
0.0029	34.8837	3000	0.0081	0.1256
0.0025	36.0465	3100	0.0081	0.1347
0.0027	37.2093	3200	0.0081	0.1324
0.0028	38.3721	3300	0.0082	0.1324
0.0023	39.5349	3400	0.0082	0.1324

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1