vit-base_rvl-cdip-tiny_rvl_cdip-NK1000_kd_MSE

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: 2.1927
Accuracy: 0.5835
Brier Loss: 0.6740
Nll: 3.1975
F1 Micro: 0.5835
F1 Macro: 0.5865
Ece: 0.2742
Aurc: 0.2074

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: 64
eval_batch_size: 64
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Brier Loss	Nll	F1 Micro	F1 Macro	Ece	Aurc
No log	1.0	250	4.2227	0.1325	0.9130	6.8924	0.1325	0.0728	0.0573	0.7519
4.2305	2.0	500	3.9645	0.1638	0.8922	5.8361	0.1638	0.1235	0.0588	0.7012
4.2305	3.0	750	3.6177	0.285	0.8227	4.3429	0.285	0.2289	0.0627	0.5424
3.6208	4.0	1000	3.2220	0.3733	0.7617	3.5860	0.3733	0.3356	0.0606	0.4322
3.6208	5.0	1250	3.0177	0.4045	0.7308	3.7807	0.4045	0.3770	0.0721	0.3835
2.9674	6.0	1500	2.8203	0.4365	0.7032	3.3569	0.4365	0.4130	0.0969	0.3443
2.9674	7.0	1750	2.6164	0.4557	0.6762	3.4281	0.4557	0.4413	0.0810	0.3058
2.5154	8.0	2000	2.4991	0.472	0.6651	3.3938	0.472	0.4524	0.1092	0.2846
2.5154	9.0	2250	2.4375	0.4878	0.6826	3.1749	0.4878	0.4603	0.1631	0.2872
2.2165	10.0	2500	2.3537	0.5018	0.6686	3.1767	0.5018	0.4855	0.1589	0.2743
2.2165	11.0	2750	2.2613	0.515	0.6276	3.1281	0.515	0.5141	0.1101	0.2457
1.9636	12.0	3000	2.2592	0.5242	0.6624	3.1164	0.5242	0.5131	0.1840	0.2515
1.9636	13.0	3250	2.1751	0.5315	0.6190	3.2643	0.5315	0.5268	0.1349	0.2288
1.7526	14.0	3500	2.2171	0.5248	0.6546	3.1179	0.5248	0.5162	0.1889	0.2537
1.7526	15.0	3750	2.1185	0.5507	0.6126	3.1117	0.5507	0.5496	0.1578	0.2219
1.5673	16.0	4000	2.0807	0.5537	0.6208	3.2624	0.5537	0.5459	0.1735	0.2151
1.5673	17.0	4250	2.0743	0.5677	0.6095	3.2650	0.5677	0.5683	0.1628	0.2090
1.3823	18.0	4500	2.1201	0.5605	0.6454	3.1499	0.5605	0.5558	0.2130	0.2316
1.3823	19.0	4750	2.0835	0.5655	0.6312	3.2920	0.5655	0.5666	0.2015	0.2149
1.2113	20.0	5000	2.0809	0.5675	0.6284	3.2923	0.5675	0.5675	0.2180	0.2047
1.2113	21.0	5250	2.1507	0.5633	0.6608	3.2713	0.5633	0.5668	0.2380	0.2183
1.0543	22.0	5500	2.1295	0.5683	0.6476	3.5120	0.5683	0.5672	0.2369	0.2105
1.0543	23.0	5750	2.1610	0.5675	0.6564	3.3818	0.5675	0.5625	0.2393	0.2166
0.9098	24.0	6000	2.0862	0.5735	0.6562	3.3228	0.5735	0.5782	0.2528	0.2047
0.9098	25.0	6250	2.0680	0.5727	0.6439	3.2971	0.5727	0.5767	0.2357	0.2050
0.7832	26.0	6500	2.1829	0.5763	0.6667	3.3547	0.5763	0.5792	0.2627	0.2084
0.7832	27.0	6750	2.1163	0.586	0.6479	3.2468	0.586	0.5894	0.2509	0.2016
0.6572	28.0	7000	2.1492	0.5715	0.6612	3.4268	0.5715	0.5780	0.2642	0.2114
0.6572	29.0	7250	2.1975	0.5723	0.6777	3.4662	0.5723	0.5739	0.2749	0.2202
0.5632	30.0	7500	2.1733	0.5693	0.6767	3.3743	0.5693	0.5745	0.2737	0.2170
0.5632	31.0	7750	2.1694	0.5807	0.6661	3.3917	0.5807	0.5814	0.2645	0.2193
0.4827	32.0	8000	2.1585	0.5805	0.6671	3.3811	0.5805	0.5812	0.2692	0.2150
0.4827	33.0	8250	2.1963	0.5767	0.6754	3.4575	0.5767	0.5835	0.2710	0.2160
0.4134	34.0	8500	2.1720	0.581	0.6694	3.3663	0.581	0.5811	0.2672	0.2131
0.4134	35.0	8750	2.1880	0.575	0.6759	3.4587	0.575	0.5790	0.2783	0.2105
0.3541	36.0	9000	2.1482	0.581	0.6628	3.2956	0.581	0.5842	0.2712	0.2056
0.3541	37.0	9250	2.1631	0.5885	0.6652	3.3217	0.5885	0.5915	0.2671	0.2069
0.3078	38.0	9500	2.2036	0.577	0.6811	3.3564	0.577	0.5803	0.2849	0.2141
0.3078	39.0	9750	2.1904	0.5753	0.6756	3.2783	0.5753	0.5765	0.2756	0.2135
0.2671	40.0	10000	2.1774	0.5775	0.6685	3.3109	0.5775	0.5813	0.2700	0.2084
0.2671	41.0	10250	2.1822	0.5807	0.6730	3.2139	0.5807	0.5842	0.2770	0.2100
0.2331	42.0	10500	2.1673	0.5817	0.6705	3.2960	0.5817	0.5864	0.2757	0.2070
0.2331	43.0	10750	2.1730	0.5765	0.6705	3.2195	0.5765	0.5807	0.2784	0.2072
0.2038	44.0	11000	2.1709	0.585	0.6649	3.1928	0.585	0.5893	0.2627	0.2055
0.2038	45.0	11250	2.1745	0.5783	0.6678	3.1900	0.5783	0.5811	0.2736	0.2061
0.1792	46.0	11500	2.1824	0.5835	0.6682	3.1909	0.5835	0.5858	0.2719	0.2070
0.1792	47.0	11750	2.1892	0.584	0.6716	3.2457	0.584	0.5864	0.2706	0.2082
0.16	48.0	12000	2.1820	0.5835	0.6716	3.2011	0.5835	0.5857	0.2743	0.2073
0.16	49.0	12250	2.1884	0.582	0.6736	3.2114	0.582	0.5856	0.2755	0.2073
0.1465	50.0	12500	2.1927	0.5835	0.6740	3.1975	0.5835	0.5865	0.2742	0.2074

Framework versions

Transformers 4.26.1
Pytorch 1.13.1.post200
Datasets 2.9.0
Tokenizers 0.13.2

jordyvl/vit-base_rvl-cdip-tiny_rvl_cdip-NK1000_kd_MSE

vit-base_rvl-cdip-tiny_rvl_cdip-NK1000_kd_MSE

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions