resnet101_rvl-cdip-cnn_rvl_cdip-NK1000_kd_CEKD_t2.5_a0.5

This model is a fine-tuned version of microsoft/resnet-50 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5837
• Accuracy: 0.7867
• Brier Loss: 0.3013
• Nll: 1.9882
• F1 Micro: 0.7868
• F1 Macro: 0.7860
• Ece: 0.0529
• Aurc: 0.0581

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: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Brier Loss	Nll	F1 Micro	F1 Macro	Ece	Aurc
No log	1.0	250	4.1958	0.1035	0.9350	9.1004	0.1035	0.0792	0.0472	0.9013
4.2322	2.0	500	4.0778	0.173	0.9251	6.5742	0.173	0.1393	0.0993	0.7501
4.2322	3.0	750	3.6484	0.339	0.8778	4.9108	0.339	0.2957	0.2172	0.5305
3.5256	4.0	1000	2.5967	0.4592	0.6991	3.3640	0.4592	0.4220	0.1274	0.3285
3.5256	5.0	1250	2.0345	0.5417	0.6078	3.0118	0.5417	0.5180	0.0976	0.2447
1.9172	6.0	1500	1.4417	0.625	0.5029	2.7890	0.625	0.6123	0.0549	0.1623
1.9172	7.0	1750	1.3298	0.639	0.4852	2.6110	0.639	0.6320	0.0558	0.1501
1.1801	8.0	2000	1.1697	0.674	0.4473	2.4787	0.674	0.6712	0.0466	0.1283
1.1801	9.0	2250	0.9625	0.7093	0.4020	2.3242	0.7093	0.7085	0.0526	0.1017
0.8029	10.0	2500	0.9477	0.7215	0.3893	2.3193	0.7215	0.7228	0.0515	0.0971
0.8029	11.0	2750	0.8527	0.7375	0.3692	2.2785	0.7375	0.7377	0.0490	0.0870
0.5717	12.0	3000	0.7377	0.7515	0.3470	2.1475	0.7515	0.7529	0.0552	0.0757
0.5717	13.0	3250	0.7309	0.7498	0.3469	2.1250	0.7498	0.7494	0.0589	0.0758
0.4414	14.0	3500	0.7165	0.7558	0.3427	2.1045	0.7558	0.7576	0.0582	0.0721
0.4414	15.0	3750	0.6865	0.7678	0.3319	2.0457	0.7678	0.7688	0.0551	0.0697
0.3691	16.0	4000	0.7002	0.7662	0.3348	2.1280	0.7663	0.7664	0.0567	0.0698
0.3691	17.0	4250	0.6896	0.7628	0.3326	2.0750	0.7628	0.7631	0.0608	0.0691
0.3214	18.0	4500	0.6666	0.7715	0.3258	2.0468	0.7715	0.7707	0.0544	0.0680
0.3214	19.0	4750	0.6735	0.7702	0.3277	2.0544	0.7702	0.7700	0.0571	0.0681
0.2914	20.0	5000	0.6607	0.772	0.3241	2.0364	0.772	0.7729	0.0525	0.0659
0.2914	21.0	5250	0.6625	0.7688	0.3217	2.0387	0.7688	0.7703	0.0455	0.0664
0.2653	22.0	5500	0.6543	0.775	0.3200	2.0560	0.775	0.7752	0.0507	0.0647
0.2653	23.0	5750	0.6409	0.7725	0.3188	2.0091	0.7725	0.7733	0.0554	0.0647
0.2482	24.0	6000	0.6452	0.7758	0.3191	2.0256	0.7758	0.7756	0.0502	0.0655
0.2482	25.0	6250	0.6401	0.7742	0.3196	2.0668	0.7742	0.7745	0.0528	0.0648
0.2354	26.0	6500	0.6316	0.775	0.3171	2.0150	0.775	0.7755	0.0555	0.0634
0.2354	27.0	6750	0.6257	0.7808	0.3147	2.0129	0.7808	0.7808	0.0503	0.0624
0.2229	28.0	7000	0.6343	0.7778	0.3144	2.0910	0.7778	0.7776	0.0510	0.0624
0.2229	29.0	7250	0.6206	0.781	0.3115	2.0399	0.7810	0.7798	0.0555	0.0606
0.2147	30.0	7500	0.6262	0.777	0.3124	2.0603	0.777	0.7772	0.0539	0.0616
0.2147	31.0	7750	0.6265	0.7788	0.3137	2.0833	0.7788	0.7777	0.0532	0.0614
0.2058	32.0	8000	0.6134	0.7815	0.3119	2.0369	0.7815	0.7815	0.0514	0.0615
0.2058	33.0	8250	0.6153	0.7772	0.3133	2.0513	0.7773	0.7772	0.0534	0.0623
0.1994	34.0	8500	0.6143	0.7853	0.3098	2.0188	0.7853	0.7857	0.0508	0.0611
0.1994	35.0	8750	0.6096	0.7827	0.3086	2.0134	0.7828	0.7828	0.0512	0.0606
0.1932	36.0	9000	0.6094	0.784	0.3067	2.0151	0.7840	0.7847	0.0471	0.0602
0.1932	37.0	9250	0.6142	0.7833	0.3111	2.0213	0.7833	0.7829	0.0542	0.0608
0.1895	38.0	9500	0.6103	0.7812	0.3094	2.0594	0.7812	0.7799	0.0529	0.0603
0.1895	39.0	9750	0.6059	0.781	0.3078	2.0386	0.7810	0.7806	0.0545	0.0607
0.1848	40.0	10000	0.6042	0.782	0.3072	2.0133	0.782	0.7824	0.0527	0.0603
0.1848	41.0	10250	0.5991	0.785	0.3043	2.0124	0.785	0.7853	0.0496	0.0594
0.1793	42.0	10500	0.6034	0.784	0.3058	2.0607	0.7840	0.7838	0.0490	0.0599
0.1793	43.0	10750	0.6047	0.7827	0.3068	2.0139	0.7828	0.7819	0.0492	0.0595
0.1768	44.0	11000	0.5982	0.785	0.3057	2.0303	0.785	0.7843	0.0473	0.0596
0.1768	45.0	11250	0.6036	0.7795	0.3087	2.0173	0.7795	0.7788	0.0549	0.0607
0.1743	46.0	11500	0.5974	0.785	0.3060	2.0026	0.785	0.7839	0.0478	0.0596
0.1743	47.0	11750	0.5996	0.782	0.3068	2.0144	0.782	0.7825	0.0480	0.0598
0.1707	48.0	12000	0.5958	0.7833	0.3079	2.0344	0.7833	0.7827	0.0500	0.0598
0.1707	49.0	12250	0.5969	0.782	0.3060	2.0162	0.782	0.7820	0.0482	0.0597
0.1683	50.0	12500	0.5933	0.784	0.3043	1.9897	0.7840	0.7836	0.0496	0.0589
0.1683	51.0	12750	0.5935	0.7833	0.3042	2.0142	0.7833	0.7829	0.0501	0.0586
0.1649	52.0	13000	0.5950	0.7847	0.3050	2.0125	0.7847	0.7851	0.0475	0.0591
0.1649	53.0	13250	0.5904	0.7837	0.3020	1.9830	0.7837	0.7837	0.0504	0.0584
0.1636	54.0	13500	0.5926	0.785	0.3042	2.0006	0.785	0.7845	0.0493	0.0588
0.1636	55.0	13750	0.5885	0.7847	0.3029	2.0025	0.7847	0.7843	0.0505	0.0585
0.1616	56.0	14000	0.5920	0.788	0.3041	2.0174	0.788	0.7878	0.0520	0.0591
0.1616	57.0	14250	0.5927	0.7863	0.3033	2.0321	0.7863	0.7858	0.0521	0.0588
0.1592	58.0	14500	0.5878	0.787	0.3017	1.9751	0.787	0.7874	0.0461	0.0584
0.1592	59.0	14750	0.5888	0.7867	0.3030	1.9996	0.7868	0.7864	0.0494	0.0582
0.1585	60.0	15000	0.5929	0.786	0.3052	2.0237	0.786	0.7857	0.0512	0.0584
0.1585	61.0	15250	0.5894	0.7865	0.3026	1.9895	0.7865	0.7864	0.0548	0.0585
0.1562	62.0	15500	0.5903	0.7873	0.3033	1.9670	0.7873	0.7870	0.0481	0.0584
0.1562	63.0	15750	0.5896	0.7853	0.3023	1.9681	0.7853	0.7850	0.0520	0.0587
0.1548	64.0	16000	0.5903	0.7847	0.3027	1.9865	0.7847	0.7846	0.0506	0.0587
0.1548	65.0	16250	0.5910	0.7853	0.3039	2.0009	0.7853	0.7849	0.0515	0.0593
0.1537	66.0	16500	0.5866	0.7883	0.3012	1.9561	0.7883	0.7881	0.0447	0.0581
0.1537	67.0	16750	0.5858	0.7867	0.3009	1.9868	0.7868	0.7861	0.0486	0.0577
0.1526	68.0	17000	0.5886	0.7867	0.3024	2.0009	0.7868	0.7862	0.0530	0.0587
0.1526	69.0	17250	0.5850	0.7863	0.3010	2.0095	0.7863	0.7860	0.0510	0.0581
0.1508	70.0	17500	0.5867	0.7865	0.3019	2.0304	0.7865	0.7861	0.0525	0.0583
0.1508	71.0	17750	0.5895	0.7857	0.3038	2.0013	0.7857	0.7853	0.0478	0.0586
0.15	72.0	18000	0.5894	0.7847	0.3025	2.0051	0.7847	0.7845	0.0500	0.0586
0.15	73.0	18250	0.5867	0.7865	0.3022	1.9634	0.7865	0.7860	0.0489	0.0582
0.149	74.0	18500	0.5888	0.7857	0.3026	1.9817	0.7857	0.7851	0.0497	0.0584
0.149	75.0	18750	0.5823	0.7885	0.2994	1.9873	0.7885	0.7880	0.0476	0.0577
0.1483	76.0	19000	0.5866	0.7853	0.3025	1.9870	0.7853	0.7849	0.0531	0.0583
0.1483	77.0	19250	0.5866	0.7867	0.3013	1.9933	0.7868	0.7862	0.0498	0.0577
0.1478	78.0	19500	0.5844	0.787	0.3010	1.9793	0.787	0.7868	0.0465	0.0579
0.1478	79.0	19750	0.5850	0.7857	0.3005	1.9856	0.7857	0.7855	0.0489	0.0580
0.1463	80.0	20000	0.5829	0.7893	0.2999	2.0003	0.7893	0.7890	0.0543	0.0578
0.1463	81.0	20250	0.5845	0.7867	0.3011	2.0178	0.7868	0.7864	0.0494	0.0580
0.1457	82.0	20500	0.5878	0.7865	0.3022	2.0108	0.7865	0.7861	0.0507	0.0583
0.1457	83.0	20750	0.5862	0.7865	0.3016	1.9996	0.7865	0.7865	0.0505	0.0585
0.1452	84.0	21000	0.5851	0.7863	0.3011	2.0002	0.7863	0.7859	0.0481	0.0582
0.1452	85.0	21250	0.5850	0.787	0.3013	1.9659	0.787	0.7867	0.0524	0.0582
0.1449	86.0	21500	0.5878	0.7867	0.3023	1.9837	0.7868	0.7866	0.0526	0.0581
0.1449	87.0	21750	0.5844	0.7873	0.3010	1.9807	0.7873	0.7865	0.0522	0.0577
0.1437	88.0	22000	0.5846	0.7877	0.3012	1.9947	0.7877	0.7869	0.0464	0.0580
0.1437	89.0	22250	0.5859	0.787	0.3016	2.0002	0.787	0.7867	0.0503	0.0581
0.143	90.0	22500	0.5838	0.7865	0.3010	1.9996	0.7865	0.7859	0.0496	0.0576
0.143	91.0	22750	0.5843	0.7837	0.3011	1.9683	0.7837	0.7834	0.0501	0.0583
0.1426	92.0	23000	0.5843	0.7873	0.3010	1.9960	0.7873	0.7870	0.0524	0.0578
0.1426	93.0	23250	0.5827	0.7847	0.3005	1.9719	0.7847	0.7844	0.0506	0.0579
0.1428	94.0	23500	0.5831	0.7865	0.3009	1.9781	0.7865	0.7862	0.0517	0.0579
0.1428	95.0	23750	0.5821	0.784	0.3001	1.9641	0.7840	0.7838	0.0505	0.0579
0.1424	96.0	24000	0.5850	0.7845	0.3020	1.9667	0.7845	0.7842	0.0526	0.0584
0.1424	97.0	24250	0.5850	0.7847	0.3012	1.9776	0.7847	0.7844	0.0508	0.0579
0.142	98.0	24500	0.5845	0.7877	0.3011	1.9745	0.7877	0.7870	0.0491	0.0579
0.142	99.0	24750	0.5834	0.7853	0.3010	1.9679	0.7853	0.7852	0.0506	0.0581
0.1416	100.0	25000	0.5837	0.7867	0.3013	1.9882	0.7868	0.7860	0.0529	0.0581

Framework versions

• Transformers 4.33.3
• Pytorch 2.2.0.dev20231002
• Datasets 2.7.1
• Tokenizers 0.13.3