finetuned-food101

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the food101 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6105
• Accuracy: 0.8400

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: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
4.1344	0.0248	100	4.0304	0.3063
3.5328	0.0497	200	3.3729	0.4410
2.9715	0.0745	300	2.8900	0.5135
2.724	0.0994	400	2.5096	0.5443
2.311	0.1242	500	2.1726	0.5895
2.266	0.1491	600	2.0223	0.5880
1.9671	0.1739	700	1.7585	0.6330
1.8617	0.1988	800	1.7300	0.6212
1.4694	0.2236	900	1.7507	0.6078
1.7876	0.2484	1000	1.6520	0.6133
1.7647	0.2733	1100	1.4576	0.6598
1.7	0.2981	1200	1.4420	0.6577
1.533	0.3230	1300	1.4389	0.6537
1.3895	0.3478	1400	1.4178	0.6587
1.5497	0.3727	1500	1.3048	0.6861
1.3327	0.3975	1600	1.3361	0.6714
1.53	0.4224	1700	1.3425	0.6697
1.538	0.4472	1800	1.3453	0.6642
1.5056	0.4720	1900	1.2742	0.6783
1.2728	0.4969	2000	1.1779	0.7045
1.1734	0.5217	2100	1.2630	0.6808
1.527	0.5466	2200	1.1810	0.7023
1.3873	0.5714	2300	1.1831	0.7040
1.3545	0.5963	2400	1.1836	0.7002
1.4842	0.6211	2500	1.1441	0.7129
1.1974	0.6460	2600	1.1230	0.7155
1.4204	0.6708	2700	1.1766	0.7002
1.152	0.6957	2800	1.2166	0.6950
1.162	0.7205	2900	1.1674	0.7003
1.4516	0.7453	3000	1.1207	0.7140
1.2378	0.7702	3100	1.2072	0.6906
0.991	0.7950	3200	1.1122	0.7131
1.3078	0.8199	3300	1.1207	0.7170
1.1483	0.8447	3400	1.0665	0.7245
1.453	0.8696	3500	1.0640	0.7267
1.4457	0.8944	3600	1.0565	0.7321
1.1636	0.9193	3700	1.0576	0.7255
1.157	0.9441	3800	1.0648	0.7261
1.1923	0.9689	3900	1.0473	0.7271
1.2325	0.9938	4000	1.0501	0.7298
1.1503	1.0186	4100	1.0566	0.7243
1.0633	1.0435	4200	1.0005	0.7444
1.2061	1.0683	4300	1.0196	0.7377
1.0315	1.0932	4400	1.0139	0.7392
1.038	1.1180	4500	1.0299	0.7318
0.7728	1.1429	4600	1.0522	0.7257
0.9302	1.1677	4700	1.0219	0.7362
1.1084	1.1925	4800	0.9940	0.7349
1.0345	1.2174	4900	0.9775	0.7446
1.0541	1.2422	5000	1.0076	0.7366
0.9345	1.2671	5100	1.0075	0.7398
0.9149	1.2919	5200	1.0558	0.7261
1.2583	1.3168	5300	0.9703	0.7476
1.0745	1.3416	5400	0.9902	0.7425
0.8319	1.3665	5500	0.9442	0.7553
1.1286	1.3913	5600	0.9620	0.7532
0.8228	1.4161	5700	0.9329	0.7555
1.3209	1.4410	5800	0.9402	0.7543
0.7629	1.4658	5900	0.9497	0.7547
0.9906	1.4907	6000	0.9362	0.7589
0.9966	1.5155	6100	0.9322	0.7595
0.8868	1.5404	6200	0.9613	0.7506
0.956	1.5652	6300	0.9370	0.7568
1.1833	1.5901	6400	0.9277	0.7597
0.9747	1.6149	6500	0.8777	0.7696
1.0119	1.6398	6600	0.8980	0.7653
0.9764	1.6646	6700	0.9071	0.7641
1.0528	1.6894	6800	0.8941	0.7694
0.942	1.7143	6900	0.8718	0.7737
1.0387	1.7391	7000	0.8615	0.7787
0.9054	1.7640	7100	0.8689	0.7735
1.0327	1.7888	7200	0.8953	0.7692
0.8425	1.8137	7300	0.8533	0.7761
0.9388	1.8385	7400	0.8772	0.7687
1.1037	1.8634	7500	0.8634	0.7731
0.9659	1.8882	7600	0.8502	0.7766
1.0133	1.9130	7700	0.8479	0.7766
0.8395	1.9379	7800	0.8052	0.7889
0.8803	1.9627	7900	0.8379	0.7775
0.7866	1.9876	8000	0.8283	0.7835
0.5067	2.0124	8100	0.8207	0.7835
0.7083	2.0373	8200	0.8320	0.7803
0.6581	2.0621	8300	0.8162	0.7869
0.7376	2.0870	8400	0.8222	0.7871
0.6492	2.1118	8500	0.8153	0.7868
0.6356	2.1366	8600	0.7930	0.7929
0.7626	2.1615	8700	0.8167	0.7874
0.7389	2.1863	8800	0.8076	0.7889
0.503	2.2112	8900	0.8312	0.7869
0.7901	2.2360	9000	0.8137	0.7900
0.8387	2.2609	9100	0.8207	0.7832
0.7048	2.2857	9200	0.8105	0.7898
0.6412	2.3106	9300	0.7829	0.7950
0.6864	2.3354	9400	0.7851	0.7941
0.7411	2.3602	9500	0.7642	0.8031
0.6221	2.3851	9600	0.7603	0.8030
0.7769	2.4099	9700	0.7846	0.7975
0.7939	2.4348	9800	0.7914	0.7933
0.5641	2.4596	9900	0.7700	0.7992
0.8009	2.4845	10000	0.7699	0.8015
0.6111	2.5093	10100	0.7603	0.8036
0.925	2.5342	10200	0.7727	0.8003
0.6206	2.5590	10300	0.7765	0.7984
0.5977	2.5839	10400	0.7793	0.7960
0.8146	2.6087	10500	0.7799	0.7978
0.7869	2.6335	10600	0.7396	0.8087
0.8966	2.6584	10700	0.7386	0.8071
0.6654	2.6832	10800	0.7305	0.8103
0.737	2.7081	10900	0.7317	0.8083
0.9283	2.7329	11000	0.7409	0.8072
0.7491	2.7578	11100	0.7088	0.8153
0.6807	2.7826	11200	0.7154	0.8123
0.4485	2.8075	11300	0.6985	0.8180
0.6694	2.8323	11400	0.7124	0.8147
0.6661	2.8571	11500	0.7075	0.8153
0.7971	2.8820	11600	0.7375	0.8078
0.9771	2.9068	11700	0.7133	0.8133
0.5238	2.9317	11800	0.7077	0.8157
0.5636	2.9565	11900	0.7419	0.8030
0.8962	2.9814	12000	0.7021	0.8175
0.4561	3.0062	12100	0.7031	0.8162
0.4906	3.0311	12200	0.7104	0.8171
0.5422	3.0559	12300	0.7035	0.8154
0.5541	3.0807	12400	0.6905	0.8232
0.5009	3.1056	12500	0.6994	0.8173
0.4567	3.1304	12600	0.6911	0.8203
0.4431	3.1553	12700	0.6933	0.8192
0.5915	3.1801	12800	0.6838	0.8221
0.5551	3.2050	12900	0.6886	0.8199
0.4528	3.2298	13000	0.6883	0.8212
0.5563	3.2547	13100	0.6867	0.8192
0.4836	3.2795	13200	0.6771	0.8253
0.4535	3.3043	13300	0.6713	0.8249
0.468	3.3292	13400	0.6616	0.8270
0.4691	3.3540	13500	0.6707	0.8261
0.4784	3.3789	13600	0.6733	0.8241
0.5187	3.4037	13700	0.6658	0.8251
0.5105	3.4286	13800	0.6631	0.8275
0.3935	3.4534	13900	0.6656	0.8283
0.463	3.4783	14000	0.6554	0.8301
0.3259	3.5031	14100	0.6640	0.8292
0.7286	3.5280	14200	0.6500	0.8308
0.4422	3.5528	14300	0.6540	0.8313
0.4374	3.5776	14400	0.6497	0.8317
0.7962	3.6025	14500	0.6416	0.8340
0.6297	3.6273	14600	0.6393	0.8339
0.4933	3.6522	14700	0.6379	0.8336
0.5548	3.6770	14800	0.6300	0.8356
0.564	3.7019	14900	0.6284	0.8352
0.2638	3.7267	15000	0.6299	0.8338
0.6129	3.7516	15100	0.6253	0.8374
0.51	3.7764	15200	0.6205	0.8390
0.4612	3.8012	15300	0.6165	0.8390
0.5304	3.8261	15400	0.6112	0.8412
0.4738	3.8509	15500	0.6149	0.8388
0.3845	3.8758	15600	0.6141	0.8391
0.4533	3.9006	15700	0.6139	0.8399
0.3539	3.9255	15800	0.6131	0.8402
0.6485	3.9503	15900	0.6118	0.8397
0.331	3.9752	16000	0.6108	0.8397
0.3582	4.0	16100	0.6105	0.8400

Framework versions

• Transformers 4.40.1
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1