Hierarchical_Agent_Action

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

• Loss: 0.5942
• Accuracy: 0.8403

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.4407	0.81	100	2.2716	0.6058
1.7756	1.61	200	1.6162	0.7065
1.3948	2.42	300	1.2200	0.7698
1.131	3.23	400	1.0012	0.7856
0.9239	4.03	500	0.9055	0.7827
0.8699	4.84	600	0.8103	0.7827
0.6707	5.65	700	0.7610	0.7842
0.6206	6.45	800	0.7312	0.7885
0.5795	7.26	900	0.6989	0.8101
0.4914	8.06	1000	0.7066	0.7813
0.5087	8.87	1100	0.6398	0.8187
0.4373	9.68	1200	0.6293	0.8043
0.4365	10.48	1300	0.6726	0.7971
0.4517	11.29	1400	0.6047	0.8245
0.4114	12.1	1500	0.6088	0.8230
0.426	12.9	1600	0.6165	0.8201
0.3456	13.71	1700	0.6133	0.8259
0.332	14.52	1800	0.6736	0.8201
0.3646	15.32	1900	0.6406	0.8173
0.3287	16.13	2000	0.6978	0.7971
0.2793	16.94	2100	0.6433	0.8173
0.2924	17.74	2200	0.6474	0.8144
0.2605	18.55	2300	0.6279	0.8288
0.2016	19.35	2400	0.6361	0.8216
0.2524	20.16	2500	0.6394	0.8259
0.2017	20.97	2600	0.6683	0.8158
0.2082	21.77	2700	0.6389	0.8345
0.2751	22.58	2800	0.6141	0.8374
0.207	23.39	2900	0.6052	0.8259
0.1791	24.19	3000	0.6332	0.8230
0.1719	25.0	3100	0.5942	0.8403
0.1685	25.81	3200	0.6121	0.8360
0.1557	26.61	3300	0.6237	0.8345
0.1694	27.42	3400	0.6372	0.8317
0.1927	28.23	3500	0.6378	0.8273
0.1375	29.03	3600	0.6258	0.8331
0.1653	29.84	3700	0.6262	0.8331

Framework versions

• Transformers 4.35.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.14.1