dit-base_tobacco-tiny_tobacco3482_kd_CEKD_t2.5_a0.5

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

• Loss: 0.6206
• Accuracy: 0.825
• Brier Loss: 0.2570
• Nll: 0.9939
• F1 Micro: 0.825
• F1 Macro: 0.8166
• Ece: 0.1370
• Aurc: 0.0444

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	4.0511	0.22	0.9320	7.5162	0.22	0.0792	0.3104	0.7615
No log	2.0	14	3.4353	0.35	0.8214	5.1797	0.35	0.2316	0.2589	0.6234
No log	3.0	21	2.6406	0.47	0.6828	2.9202	0.47	0.3725	0.2781	0.3277
No log	4.0	28	2.0027	0.57	0.5596	1.8624	0.57	0.4971	0.2526	0.2124
No log	5.0	35	1.5018	0.65	0.4518	1.7094	0.65	0.6128	0.2242	0.1396
No log	6.0	42	1.3904	0.71	0.4105	1.8765	0.7100	0.7058	0.2202	0.1126
No log	7.0	49	1.1226	0.76	0.3558	1.8024	0.76	0.7029	0.1815	0.0841
No log	8.0	56	1.1810	0.73	0.3716	1.5642	0.7300	0.7027	0.1956	0.0834
No log	9.0	63	1.2131	0.73	0.3811	1.7544	0.7300	0.6774	0.2070	0.0872
No log	10.0	70	1.3986	0.72	0.4043	2.0161	0.72	0.7259	0.2021	0.1098
No log	11.0	77	1.1001	0.765	0.3202	1.9113	0.765	0.7578	0.1859	0.0678
No log	12.0	84	1.0429	0.77	0.3487	1.2955	0.7700	0.7663	0.1910	0.0827
No log	13.0	91	0.9864	0.77	0.3227	1.3721	0.7700	0.7734	0.1692	0.0710
No log	14.0	98	1.0068	0.74	0.3581	1.3362	0.74	0.7271	0.1848	0.0804
No log	15.0	105	0.8635	0.795	0.3009	1.4785	0.795	0.7810	0.1646	0.0538
No log	16.0	112	0.8157	0.81	0.2845	1.2525	0.81	0.7931	0.1545	0.0519
No log	17.0	119	0.8616	0.78	0.3186	1.4230	0.78	0.7705	0.1610	0.0647
No log	18.0	126	0.8034	0.8	0.2784	1.4410	0.8000	0.7811	0.1576	0.0489
No log	19.0	133	0.7601	0.805	0.2697	1.2885	0.805	0.7823	0.1499	0.0494
No log	20.0	140	0.7598	0.82	0.2709	1.3643	0.82	0.8090	0.1542	0.0516
No log	21.0	147	0.8221	0.79	0.2905	1.4031	0.79	0.7640	0.1612	0.0585
No log	22.0	154	0.7271	0.825	0.2599	1.0950	0.825	0.8147	0.1381	0.0454
No log	23.0	161	0.7556	0.795	0.2891	1.1111	0.795	0.7822	0.1413	0.0558
No log	24.0	168	0.7197	0.81	0.2759	1.1361	0.81	0.7905	0.1617	0.0500
No log	25.0	175	0.7192	0.83	0.2620	1.3395	0.83	0.8155	0.1459	0.0433
No log	26.0	182	0.7347	0.805	0.2821	1.1396	0.805	0.7868	0.1512	0.0541
No log	27.0	189	0.7402	0.815	0.2805	1.3562	0.815	0.7928	0.1489	0.0519
No log	28.0	196	0.6986	0.815	0.2562	1.1454	0.815	0.7944	0.1467	0.0443
No log	29.0	203	0.7148	0.81	0.2718	1.1404	0.81	0.7944	0.1440	0.0513
No log	30.0	210	0.7041	0.81	0.2796	1.3773	0.81	0.7998	0.1484	0.0494
No log	31.0	217	0.7428	0.815	0.2823	1.1146	0.815	0.7967	0.1626	0.0542
No log	32.0	224	0.6941	0.82	0.2682	1.1921	0.82	0.8098	0.1427	0.0478
No log	33.0	231	0.7170	0.81	0.2794	1.2244	0.81	0.7875	0.1407	0.0511
No log	34.0	238	0.7024	0.815	0.2805	1.0423	0.815	0.8043	0.1560	0.0512
No log	35.0	245	0.7299	0.81	0.2710	1.1835	0.81	0.7964	0.1475	0.0530
No log	36.0	252	0.6488	0.83	0.2500	1.1662	0.83	0.8117	0.1315	0.0431
No log	37.0	259	0.6877	0.815	0.2751	1.0878	0.815	0.7973	0.1381	0.0489
No log	38.0	266	0.7019	0.84	0.2620	1.2709	0.8400	0.8282	0.1607	0.0498
No log	39.0	273	0.6687	0.81	0.2680	1.3004	0.81	0.7959	0.1346	0.0465
No log	40.0	280	0.6813	0.81	0.2809	1.0539	0.81	0.7929	0.1628	0.0500
No log	41.0	287	0.6525	0.83	0.2493	1.1496	0.83	0.8176	0.1413	0.0437
No log	42.0	294	0.6526	0.835	0.2547	1.2429	0.835	0.8253	0.1420	0.0450
No log	43.0	301	0.6696	0.82	0.2717	1.0446	0.82	0.8118	0.1486	0.0501
No log	44.0	308	0.6555	0.83	0.2626	0.9948	0.83	0.8214	0.1366	0.0461
No log	45.0	315	0.6380	0.82	0.2600	1.2151	0.82	0.8026	0.1263	0.0428
No log	46.0	322	0.6356	0.82	0.2571	1.0923	0.82	0.8114	0.1443	0.0449
No log	47.0	329	0.6444	0.815	0.2638	1.0657	0.815	0.7980	0.1503	0.0476
No log	48.0	336	0.6337	0.82	0.2676	1.0650	0.82	0.8077	0.1370	0.0442
No log	49.0	343	0.6271	0.84	0.2541	1.1500	0.8400	0.8230	0.1365	0.0422
No log	50.0	350	0.6284	0.81	0.2588	1.2703	0.81	0.7964	0.1411	0.0425
No log	51.0	357	0.6507	0.82	0.2612	1.1306	0.82	0.7996	0.1558	0.0460
No log	52.0	364	0.6329	0.825	0.2602	1.2060	0.825	0.8146	0.1296	0.0439
No log	53.0	371	0.6342	0.825	0.2574	1.0132	0.825	0.8158	0.1467	0.0434
No log	54.0	378	0.6486	0.82	0.2633	1.1662	0.82	0.8060	0.1445	0.0466
No log	55.0	385	0.6245	0.825	0.2588	1.1358	0.825	0.8088	0.1428	0.0429
No log	56.0	392	0.6303	0.815	0.2616	0.9843	0.815	0.8013	0.1447	0.0458
No log	57.0	399	0.6196	0.82	0.2545	1.1936	0.82	0.8076	0.1516	0.0438
No log	58.0	406	0.6241	0.82	0.2620	1.0557	0.82	0.8100	0.1423	0.0450
No log	59.0	413	0.6278	0.82	0.2579	1.0777	0.82	0.8076	0.1382	0.0451
No log	60.0	420	0.6385	0.81	0.2651	0.9962	0.81	0.7910	0.1565	0.0467
No log	61.0	427	0.6328	0.82	0.2619	0.9968	0.82	0.8103	0.1299	0.0469
No log	62.0	434	0.6195	0.82	0.2571	0.9997	0.82	0.8062	0.1471	0.0438
No log	63.0	441	0.6150	0.825	0.2560	1.0061	0.825	0.8166	0.1498	0.0430
No log	64.0	448	0.6201	0.825	0.2574	1.0592	0.825	0.8166	0.1369	0.0442
No log	65.0	455	0.6281	0.815	0.2601	0.9990	0.815	0.8013	0.1449	0.0459
No log	66.0	462	0.6232	0.825	0.2538	1.0657	0.825	0.8166	0.1341	0.0442
No log	67.0	469	0.6242	0.82	0.2567	1.0622	0.82	0.8100	0.1432	0.0445
No log	68.0	476	0.6213	0.82	0.2598	1.0666	0.82	0.8100	0.1517	0.0447
No log	69.0	483	0.6268	0.82	0.2577	1.0106	0.82	0.8100	0.1365	0.0455
No log	70.0	490	0.6252	0.82	0.2579	0.9979	0.82	0.8100	0.1395	0.0451
No log	71.0	497	0.6251	0.82	0.2589	1.0606	0.82	0.8100	0.1485	0.0448
0.3286	72.0	504	0.6212	0.825	0.2571	1.0034	0.825	0.8166	0.1448	0.0443
0.3286	73.0	511	0.6212	0.82	0.2584	0.9940	0.82	0.8100	0.1499	0.0444
0.3286	74.0	518	0.6214	0.82	0.2576	0.9914	0.82	0.8100	0.1411	0.0448
0.3286	75.0	525	0.6233	0.82	0.2580	0.9966	0.82	0.8100	0.1592	0.0450
0.3286	76.0	532	0.6214	0.82	0.2568	0.9952	0.82	0.8100	0.1404	0.0448
0.3286	77.0	539	0.6217	0.825	0.2575	0.9951	0.825	0.8166	0.1361	0.0445
0.3286	78.0	546	0.6220	0.82	0.2569	0.9964	0.82	0.8100	0.1385	0.0450
0.3286	79.0	553	0.6225	0.82	0.2581	0.9950	0.82	0.8100	0.1485	0.0450
0.3286	80.0	560	0.6213	0.82	0.2578	0.9912	0.82	0.8100	0.1381	0.0446
0.3286	81.0	567	0.6209	0.82	0.2572	0.9948	0.82	0.8100	0.1415	0.0447
0.3286	82.0	574	0.6213	0.82	0.2578	0.9958	0.82	0.8100	0.1422	0.0449
0.3286	83.0	581	0.6220	0.82	0.2579	0.9947	0.82	0.8100	0.1553	0.0448
0.3286	84.0	588	0.6212	0.82	0.2574	0.9915	0.82	0.8100	0.1418	0.0447
0.3286	85.0	595	0.6220	0.82	0.2579	0.9937	0.82	0.8100	0.1628	0.0450
0.3286	86.0	602	0.6207	0.82	0.2572	0.9945	0.82	0.8100	0.1412	0.0447
0.3286	87.0	609	0.6212	0.82	0.2573	0.9940	0.82	0.8100	0.1414	0.0447
0.3286	88.0	616	0.6201	0.825	0.2570	0.9943	0.825	0.8166	0.1366	0.0443
0.3286	89.0	623	0.6210	0.82	0.2573	0.9944	0.82	0.8100	0.1414	0.0448
0.3286	90.0	630	0.6207	0.82	0.2572	0.9942	0.82	0.8100	0.1414	0.0447
0.3286	91.0	637	0.6210	0.82	0.2572	0.9952	0.82	0.8100	0.1415	0.0447
0.3286	92.0	644	0.6205	0.82	0.2572	0.9939	0.82	0.8100	0.1414	0.0447
0.3286	93.0	651	0.6207	0.825	0.2570	0.9938	0.825	0.8166	0.1373	0.0445
0.3286	94.0	658	0.6206	0.82	0.2572	0.9945	0.82	0.8100	0.1414	0.0447
0.3286	95.0	665	0.6203	0.825	0.2568	0.9951	0.825	0.8166	0.1370	0.0444
0.3286	96.0	672	0.6205	0.82	0.2571	0.9942	0.82	0.8100	0.1413	0.0448
0.3286	97.0	679	0.6206	0.825	0.2570	0.9943	0.825	0.8166	0.1370	0.0445
0.3286	98.0	686	0.6206	0.825	0.2570	0.9942	0.825	0.8166	0.1370	0.0445
0.3286	99.0	693	0.6206	0.825	0.2570	0.9940	0.825	0.8166	0.1370	0.0445
0.3286	100.0	700	0.6206	0.825	0.2570	0.9939	0.825	0.8166	0.1370	0.0444

Framework versions

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