invitrace-vit-base-food

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

• Loss: 1.7567
• Accuracy: 0.6960

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: 2e-05
• 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
5.0781	0.0803	100	5.0829	0.0433
4.9949	0.1605	200	4.9850	0.0857
4.9031	0.2408	300	4.8898	0.1674
4.7506	0.3210	400	4.7910	0.1959
4.718	0.4013	500	4.6958	0.2248
4.5883	0.4815	600	4.6033	0.2378
4.5255	0.5618	700	4.5165	0.2846
4.4352	0.6421	800	4.4288	0.3209
4.3072	0.7223	900	4.3482	0.3325
4.2581	0.8026	1000	4.2725	0.3693
4.1635	0.8828	1100	4.1993	0.3701
4.2149	0.9631	1200	4.1301	0.4036
3.9068	1.0433	1300	4.0589	0.4212
3.906	1.1236	1400	3.9937	0.4258
3.8074	1.2039	1500	3.9265	0.4401
3.846	1.2841	1600	3.8624	0.4636
3.7287	1.3644	1700	3.7921	0.4746
3.6785	1.4446	1800	3.7291	0.4939
3.6254	1.5249	1900	3.6768	0.5073
3.5377	1.6051	2000	3.6177	0.5001
3.4988	1.6854	2100	3.5622	0.5150
3.5051	1.7657	2200	3.5017	0.5133
3.4085	1.8459	2300	3.4415	0.5238
3.3978	1.9262	2400	3.3934	0.5348
3.3711	2.0064	2500	3.3364	0.5360
3.1439	2.0867	2600	3.2798	0.5449
3.1757	2.1669	2700	3.2288	0.5509
2.9329	2.2472	2800	3.1791	0.5521
2.8305	2.3274	2900	3.1239	0.5671
2.8984	2.4077	3000	3.0785	0.5609
3.0165	2.4880	3100	3.0352	0.5697
2.8105	2.5682	3200	2.9885	0.5820
2.7716	2.6485	3300	2.9396	0.5846
2.8448	2.7287	3400	2.8953	0.5884
2.6256	2.8090	3500	2.8594	0.5866
2.658	2.8892	3600	2.8157	0.5970
2.5392	2.9695	3700	2.7728	0.6008
2.4969	3.0498	3800	2.7359	0.6091
2.2975	3.1300	3900	2.6909	0.6187
2.1984	3.2103	4000	2.6527	0.6181
2.2088	3.2905	4100	2.6191	0.6217
2.147	3.3708	4200	2.5753	0.6235
2.0769	3.4510	4300	2.5453	0.6241
2.3726	3.5313	4400	2.5109	0.6362
2.0648	3.6116	4500	2.4840	0.6245
2.1653	3.6918	4600	2.4469	0.6366
1.8567	3.7721	4700	2.4177	0.6398
2.1076	3.8523	4800	2.3938	0.6386
2.0246	3.9326	4900	2.3539	0.6480
1.8459	4.0128	5000	2.3342	0.6432
1.7608	4.0931	5100	2.3074	0.6476
1.7977	4.1734	5200	2.2793	0.6516
1.7611	4.2536	5300	2.2519	0.6526
1.777	4.3339	5400	2.2320	0.6512
1.642	4.4141	5500	2.2008	0.6594
1.7745	4.4944	5600	2.1775	0.6592
1.6992	4.5746	5700	2.1538	0.6604
1.6188	4.6549	5800	2.1346	0.6627
1.6404	4.7352	5900	2.1158	0.6705
1.6618	4.8154	6000	2.0963	0.6671
1.4585	4.8957	6100	2.0718	0.6705
1.5439	4.9759	6200	2.0525	0.6719
1.3129	5.0562	6300	2.0362	0.6753
1.4422	5.1364	6400	2.0254	0.6723
1.3851	5.2167	6500	2.0113	0.6739
1.5312	5.2970	6600	1.9929	0.6785
1.4072	5.3772	6700	1.9740	0.6781
1.4402	5.4575	6800	1.9656	0.6721
1.4257	5.5377	6900	1.9462	0.6827
1.3762	5.6180	7000	1.9337	0.6853
1.2985	5.6982	7100	1.9214	0.6779
1.2026	5.7785	7200	1.9081	0.6831
1.2968	5.8587	7300	1.8945	0.6847
1.3511	5.9390	7400	1.8835	0.6901
1.1626	6.0193	7500	1.8732	0.6867
1.1289	6.0995	7600	1.8625	0.6903
1.1951	6.1798	7700	1.8548	0.6847
1.1203	6.2600	7800	1.8456	0.6893
1.0551	6.3403	7900	1.8386	0.6885
0.9919	6.4205	8000	1.8277	0.6914
1.1204	6.5008	8100	1.8207	0.6893
1.2432	6.5811	8200	1.8146	0.6895
1.0495	6.6613	8300	1.8016	0.6920
1.0898	6.7416	8400	1.8014	0.6911
1.1495	6.8218	8500	1.7988	0.6948
1.2172	6.9021	8600	1.7952	0.6895
1.063	6.9823	8700	1.7848	0.6968
1.0807	7.0626	8800	1.7835	0.6926
0.9908	7.1429	8900	1.7796	0.6944
1.0848	7.2231	9000	1.7796	0.6946
1.0682	7.3034	9100	1.7703	0.6922
0.9353	7.3836	9200	1.7686	0.6948
1.0604	7.4639	9300	1.7650	0.6932
0.9961	7.5441	9400	1.7639	0.6938
1.096	7.6244	9500	1.7624	0.6964
0.9436	7.7047	9600	1.7599	0.6952
1.0565	7.7849	9700	1.7591	0.6954
0.9172	7.8652	9800	1.7579	0.6958
1.0549	7.9454	9900	1.7567	0.6960

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1