exper6_mesum5

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the sudo-s/herbier_mesuem5 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8241
• Accuracy: 0.8036

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: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.9276	0.23	100	3.8550	0.2089
3.0853	0.47	200	3.1106	0.3414
2.604	0.7	300	2.5732	0.4379
2.3183	0.93	400	2.2308	0.4882
1.5326	1.16	500	1.7903	0.5828
1.3367	1.4	600	1.5524	0.6349
1.1544	1.63	700	1.3167	0.6645
1.0788	1.86	800	1.3423	0.6385
0.6762	2.09	900	1.0780	0.7124
0.6483	2.33	1000	1.0090	0.7284
0.6321	2.56	1100	1.0861	0.7024
0.5558	2.79	1200	0.9933	0.7183
0.342	3.02	1300	0.8871	0.7462
0.2964	3.26	1400	0.9330	0.7408
0.1959	3.49	1500	0.9367	0.7343
0.368	3.72	1600	0.8472	0.7550
0.1821	3.95	1700	0.8937	0.7568
0.1851	4.19	1800	0.9546	0.7485
0.1648	4.42	1900	0.9790	0.7355
0.172	4.65	2000	0.8947	0.7627
0.0928	4.88	2100	1.0093	0.7462
0.0699	5.12	2200	0.8374	0.7639
0.0988	5.35	2300	0.9189	0.7645
0.0822	5.58	2400	0.9512	0.7580
0.1223	5.81	2500	1.0809	0.7349
0.0509	6.05	2600	0.9297	0.7769
0.0511	6.28	2700	0.8981	0.7822
0.0596	6.51	2800	0.9468	0.7704
0.0494	6.74	2900	0.9045	0.7870
0.0643	6.98	3000	1.1559	0.7391
0.0158	7.21	3100	0.8450	0.7899
0.0129	7.44	3200	0.8241	0.8036
0.0441	7.67	3300	0.9679	0.7751
0.0697	7.91	3400	1.0387	0.7751
0.0084	8.14	3500	0.9441	0.7947
0.0182	8.37	3600	0.8967	0.7994
0.0042	8.6	3700	0.8750	0.8041
0.0028	8.84	3800	0.9349	0.8041
0.0053	9.07	3900	0.9403	0.7982
0.0266	9.3	4000	0.9966	0.7959
0.0022	9.53	4100	0.9472	0.8018
0.0018	9.77	4200	0.8717	0.8136
0.0018	10.0	4300	0.8964	0.8083
0.0046	10.23	4400	0.8623	0.8160
0.0037	10.47	4500	0.8762	0.8172
0.0013	10.7	4600	0.9028	0.8142
0.0013	10.93	4700	0.9084	0.8178
0.0013	11.16	4800	0.8733	0.8213
0.001	11.4	4900	0.8823	0.8207
0.0009	11.63	5000	0.8769	0.8213
0.0282	11.86	5100	0.8791	0.8219
0.001	12.09	5200	0.8673	0.8249
0.0016	12.33	5300	0.8633	0.8225
0.0008	12.56	5400	0.8766	0.8195
0.0008	12.79	5500	0.8743	0.8225
0.0008	13.02	5600	0.8752	0.8231
0.0008	13.26	5700	0.8676	0.8237
0.0007	13.49	5800	0.8677	0.8237
0.0008	13.72	5900	0.8703	0.8237
0.0007	13.95	6000	0.8725	0.8237
0.0006	14.19	6100	0.8741	0.8231
0.0006	14.42	6200	0.8758	0.8237
0.0008	14.65	6300	0.8746	0.8243
0.0007	14.88	6400	0.8759	0.8243
0.0007	15.12	6500	0.8803	0.8231
0.0007	15.35	6600	0.8808	0.8237
0.0007	15.58	6700	0.8798	0.8243
0.0007	15.81	6800	0.8805	0.8243

Framework versions

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