README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: exper3_mesum5
    results: []

exper3_mesum5

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6397
• Accuracy: 0.8385

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: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.895	0.23	100	3.8276	0.1935
3.1174	0.47	200	3.1217	0.3107
2.6	0.7	300	2.5399	0.4207
2.256	0.93	400	2.1767	0.5160
1.5441	1.16	500	1.8086	0.5852
1.3834	1.4	600	1.5565	0.6325
1.1995	1.63	700	1.3339	0.6763
1.0845	1.86	800	1.3299	0.6533
0.6472	2.09	900	1.0679	0.7219
0.5948	2.33	1000	1.0286	0.7124
0.5565	2.56	1100	0.9595	0.7284
0.4879	2.79	1200	0.8915	0.7420
0.2816	3.02	1300	0.8159	0.7763
0.2412	3.26	1400	0.7766	0.7911
0.2015	3.49	1500	0.7850	0.7828
0.274	3.72	1600	0.7361	0.7935
0.1244	3.95	1700	0.7299	0.7911
0.0794	4.19	1800	0.7441	0.7846
0.0915	4.42	1900	0.7614	0.7941
0.0817	4.65	2000	0.7310	0.8012
0.0561	4.88	2100	0.7222	0.8065
0.0165	5.12	2200	0.7515	0.8059
0.0168	5.35	2300	0.6687	0.8213
0.0212	5.58	2400	0.6671	0.8249
0.0389	5.81	2500	0.6893	0.8278
0.0087	6.05	2600	0.6839	0.8260
0.0087	6.28	2700	0.6412	0.8320
0.0077	6.51	2800	0.6366	0.8367
0.0065	6.74	2900	0.6697	0.8272
0.0061	6.98	3000	0.6510	0.8349
0.0185	7.21	3100	0.6452	0.8367
0.0059	7.44	3200	0.6426	0.8379
0.0062	7.67	3300	0.6398	0.8379
0.0315	7.91	3400	0.6397	0.8385

Framework versions

• Transformers 4.20.1
• Pytorch 1.12.0+cu113
• Datasets 2.3.2
• Tokenizers 0.12.1