dit-base_tobacco-small_tobacco3482_kd_NKD_t1.0_g1.5

This model is a fine-tuned version of WinKawaks/vit-small-patch16-224 on the None dataset. It achieves the following results on the evaluation set:

Loss: 3.1084
Accuracy: 0.825
Brier Loss: 0.2907
Nll: 1.2013
F1 Micro: 0.825
F1 Macro: 0.8171
Ece: 0.1500
Aurc: 0.0459

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: 128
eval_batch_size: 128
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	7	5.5631	0.135	0.9164	5.3726	0.135	0.1126	0.2543	0.8397
No log	2.0	14	4.9048	0.35	0.8238	3.0911	0.35	0.2637	0.3348	0.6709
No log	3.0	21	4.1439	0.49	0.6650	1.8580	0.49	0.4532	0.2990	0.2902
No log	4.0	28	3.5518	0.66	0.4867	1.6397	0.66	0.6303	0.2902	0.1489
No log	5.0	35	3.3371	0.755	0.3981	1.6213	0.755	0.7261	0.2670	0.0984
No log	6.0	42	3.4978	0.69	0.4211	1.5668	0.69	0.6792	0.2240	0.1170
No log	7.0	49	3.0945	0.795	0.3094	1.5507	0.795	0.7653	0.1765	0.0622
No log	8.0	56	3.0882	0.775	0.3056	1.5470	0.775	0.7500	0.1826	0.0634
No log	9.0	63	3.1861	0.745	0.3331	1.6432	0.745	0.7362	0.1822	0.0754
No log	10.0	70	2.9849	0.81	0.2789	1.5850	0.81	0.7802	0.1559	0.0548
No log	11.0	77	3.0131	0.795	0.3012	1.4820	0.795	0.7720	0.1627	0.0567
No log	12.0	84	2.9054	0.795	0.2734	1.4141	0.795	0.7843	0.1501	0.0535
No log	13.0	91	2.9704	0.815	0.2720	1.4241	0.815	0.8144	0.1584	0.0536
No log	14.0	98	2.9393	0.815	0.2627	1.4735	0.815	0.7902	0.1582	0.0504
No log	15.0	105	3.0346	0.805	0.2963	1.3649	0.805	0.7973	0.1617	0.0564
No log	16.0	112	2.9648	0.79	0.2839	1.6270	0.79	0.7722	0.1418	0.0525
No log	17.0	119	3.0458	0.82	0.2960	1.3476	0.82	0.8048	0.1575	0.0622
No log	18.0	126	2.8571	0.82	0.2754	1.3958	0.82	0.8081	0.1482	0.0493
No log	19.0	133	2.9429	0.775	0.2971	1.4302	0.775	0.7617	0.1616	0.0575
No log	20.0	140	2.8274	0.825	0.2698	1.3759	0.825	0.8081	0.1520	0.0449
No log	21.0	147	2.8769	0.81	0.2713	1.3604	0.81	0.8086	0.1390	0.0466
No log	22.0	154	2.8787	0.805	0.2694	1.3016	0.805	0.7975	0.1522	0.0435
No log	23.0	161	2.8771	0.825	0.2646	1.4753	0.825	0.8215	0.1414	0.0485
No log	24.0	168	2.8950	0.805	0.2774	1.2783	0.805	0.7754	0.1406	0.0495
No log	25.0	175	2.9780	0.825	0.2829	1.3207	0.825	0.8332	0.1402	0.0496
No log	26.0	182	2.8906	0.82	0.2653	1.3097	0.82	0.8007	0.1380	0.0454
No log	27.0	189	2.9385	0.82	0.2778	1.3039	0.82	0.8211	0.1489	0.0469
No log	28.0	196	2.8644	0.83	0.2618	1.4004	0.83	0.8325	0.1358	0.0494
No log	29.0	203	2.8761	0.82	0.2720	1.2220	0.82	0.8192	0.1411	0.0463
No log	30.0	210	2.8594	0.83	0.2620	1.3323	0.83	0.8130	0.1257	0.0448
No log	31.0	217	2.8946	0.825	0.2658	1.3388	0.825	0.8236	0.1322	0.0427
No log	32.0	224	2.8698	0.825	0.2712	1.3141	0.825	0.8107	0.1467	0.0473
No log	33.0	231	2.8106	0.83	0.2563	1.3750	0.83	0.8178	0.1126	0.0422
No log	34.0	238	2.9752	0.8	0.2881	1.3007	0.8000	0.7902	0.1522	0.0499
No log	35.0	245	2.8919	0.815	0.2886	1.3057	0.815	0.8149	0.1472	0.0468
No log	36.0	252	2.8863	0.81	0.2833	1.1973	0.81	0.8006	0.1453	0.0458
No log	37.0	259	2.8283	0.845	0.2685	1.2743	0.845	0.8438	0.1481	0.0451
No log	38.0	266	2.9174	0.815	0.2825	1.2658	0.815	0.7965	0.1408	0.0530
No log	39.0	273	2.8837	0.82	0.2775	1.2946	0.82	0.8050	0.1440	0.0472
No log	40.0	280	2.8585	0.835	0.2654	1.2830	0.835	0.8169	0.1450	0.0467
No log	41.0	287	2.9323	0.82	0.2809	1.2833	0.82	0.8085	0.1342	0.0490
No log	42.0	294	2.9525	0.82	0.2847	1.2331	0.82	0.8055	0.1352	0.0481
No log	43.0	301	2.9005	0.83	0.2819	1.2643	0.83	0.8225	0.1548	0.0482
No log	44.0	308	2.8388	0.83	0.2634	1.2662	0.83	0.8152	0.1286	0.0460
No log	45.0	315	2.8962	0.82	0.2752	1.3291	0.82	0.8127	0.1442	0.0496
No log	46.0	322	2.9479	0.815	0.2883	1.2433	0.815	0.7968	0.1540	0.0523
No log	47.0	329	2.8795	0.825	0.2737	1.2477	0.825	0.8260	0.1295	0.0447
No log	48.0	336	2.9872	0.815	0.2992	1.2556	0.815	0.8029	0.1379	0.0510
No log	49.0	343	2.8156	0.84	0.2536	1.2715	0.8400	0.8263	0.1240	0.0422
No log	50.0	350	2.9534	0.81	0.2924	1.3383	0.81	0.7937	0.1471	0.0478
No log	51.0	357	2.8604	0.855	0.2549	1.2566	0.855	0.8547	0.1318	0.0411
No log	52.0	364	2.9769	0.825	0.2828	1.2325	0.825	0.8160	0.1407	0.0480
No log	53.0	371	2.8717	0.84	0.2635	1.2511	0.8400	0.8342	0.1254	0.0434
No log	54.0	378	2.9313	0.825	0.2704	1.2676	0.825	0.8159	0.1310	0.0477
No log	55.0	385	2.8552	0.82	0.2638	1.2417	0.82	0.8031	0.1490	0.0435
No log	56.0	392	2.9680	0.845	0.2729	1.2530	0.845	0.8414	0.1349	0.0452
No log	57.0	399	2.9440	0.83	0.2796	1.2344	0.83	0.8222	0.1367	0.0450
No log	58.0	406	3.0577	0.815	0.2913	1.2232	0.815	0.8068	0.1447	0.0488
No log	59.0	413	2.8861	0.835	0.2643	1.2618	0.835	0.8280	0.1354	0.0422
No log	60.0	420	3.0007	0.825	0.2822	1.2352	0.825	0.8136	0.1342	0.0449
No log	61.0	427	2.9368	0.835	0.2746	1.2437	0.835	0.8258	0.1402	0.0437
No log	62.0	434	2.9202	0.835	0.2709	1.2281	0.835	0.8258	0.1435	0.0435
No log	63.0	441	2.9720	0.835	0.2768	1.2129	0.835	0.8354	0.1444	0.0460
No log	64.0	448	2.9993	0.835	0.2815	1.2250	0.835	0.8245	0.1526	0.0451
No log	65.0	455	2.9628	0.83	0.2725	1.2477	0.83	0.8190	0.1405	0.0439
No log	66.0	462	3.0418	0.825	0.2863	1.2244	0.825	0.8142	0.1447	0.0473
No log	67.0	469	3.0196	0.83	0.2797	1.2317	0.83	0.8223	0.1450	0.0463
No log	68.0	476	3.0227	0.835	0.2834	1.2362	0.835	0.8270	0.1416	0.0446
No log	69.0	483	3.0343	0.835	0.2837	1.2377	0.835	0.8310	0.1423	0.0455
No log	70.0	490	2.9982	0.835	0.2755	1.2247	0.835	0.8245	0.1306	0.0443
No log	71.0	497	3.0230	0.825	0.2860	1.2302	0.825	0.8171	0.1376	0.0464
2.5595	72.0	504	3.0254	0.83	0.2843	1.2190	0.83	0.8222	0.1386	0.0463
2.5595	73.0	511	3.0295	0.825	0.2851	1.2206	0.825	0.8192	0.1417	0.0462
2.5595	74.0	518	3.0381	0.83	0.2845	1.2130	0.83	0.8243	0.1423	0.0457
2.5595	75.0	525	3.0258	0.825	0.2837	1.2210	0.825	0.8171	0.1431	0.0460
2.5595	76.0	532	3.0694	0.825	0.2886	1.2091	0.825	0.8171	0.1533	0.0476
2.5595	77.0	539	3.0924	0.825	0.2939	1.2130	0.825	0.8171	0.1515	0.0473
2.5595	78.0	546	3.0956	0.82	0.2921	1.2081	0.82	0.8140	0.1539	0.0482
2.5595	79.0	553	3.0859	0.825	0.2884	1.2109	0.825	0.8220	0.1480	0.0468
2.5595	80.0	560	3.0740	0.825	0.2894	1.2081	0.825	0.8136	0.1399	0.0459
2.5595	81.0	567	3.0776	0.825	0.2901	1.2066	0.825	0.8171	0.1502	0.0462
2.5595	82.0	574	3.0736	0.83	0.2869	1.2100	0.83	0.8251	0.1405	0.0462
2.5595	83.0	581	3.0943	0.825	0.2919	1.2065	0.825	0.8171	0.1503	0.0464
2.5595	84.0	588	3.0857	0.825	0.2908	1.2080	0.825	0.8171	0.1456	0.0461
2.5595	85.0	595	3.0874	0.825	0.2890	1.2063	0.825	0.8171	0.1457	0.0461
2.5595	86.0	602	3.0863	0.825	0.2880	1.2069	0.825	0.8171	0.1453	0.0459
2.5595	87.0	609	3.0844	0.825	0.2882	1.2059	0.825	0.8171	0.1457	0.0456
2.5595	88.0	616	3.1011	0.825	0.2909	1.2034	0.825	0.8171	0.1557	0.0462
2.5595	89.0	623	3.1033	0.825	0.2912	1.2033	0.825	0.8171	0.1528	0.0463
2.5595	90.0	630	3.1004	0.825	0.2903	1.2029	0.825	0.8171	0.1541	0.0461
2.5595	91.0	637	3.0998	0.825	0.2900	1.2033	0.825	0.8171	0.1499	0.0459
2.5595	92.0	644	3.1039	0.825	0.2904	1.2023	0.825	0.8171	0.1535	0.0460
2.5595	93.0	651	3.1058	0.825	0.2906	1.2020	0.825	0.8171	0.1498	0.0460
2.5595	94.0	658	3.1057	0.825	0.2906	1.2022	0.825	0.8171	0.1504	0.0459
2.5595	95.0	665	3.1066	0.825	0.2908	1.2018	0.825	0.8171	0.1509	0.0460
2.5595	96.0	672	3.1069	0.825	0.2906	1.2018	0.825	0.8171	0.1506	0.0459
2.5595	97.0	679	3.1079	0.825	0.2906	1.2013	0.825	0.8171	0.1497	0.0459
2.5595	98.0	686	3.1085	0.825	0.2907	1.2013	0.825	0.8171	0.1500	0.0459
2.5595	99.0	693	3.1083	0.825	0.2907	1.2013	0.825	0.8171	0.1499	0.0460
2.5595	100.0	700	3.1084	0.825	0.2907	1.2013	0.825	0.8171	0.1500	0.0459

Framework versions

Transformers 4.36.0.dev0
Pytorch 2.2.0.dev20231112+cu118
Datasets 2.14.5
Tokenizers 0.14.1

jordyvl/dit-base_tobacco-small_tobacco3482_kd_NKD_t1.0_g1.5

dit-base_tobacco-small_tobacco3482_kd_NKD_t1.0_g1.5

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions