6_e_200-tiny_tobacco3482_kd_CEKD_t5.0_a0.9

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: 0.5583
Accuracy: 0.82
Brier Loss: 0.2563
Nll: 1.8898
F1 Micro: 0.82
F1 Macro: 0.8009
Ece: 0.1578
Aurc: 0.0530

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: 32
eval_batch_size: 32
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	25	1.9764	0.23	0.8621	4.6756	0.23	0.1902	0.2733	0.7604
No log	2.0	50	1.2764	0.535	0.5973	2.7212	0.535	0.4337	0.2769	0.2592
No log	3.0	75	0.9774	0.68	0.4478	2.1874	0.68	0.5915	0.2144	0.1334
No log	4.0	100	0.8047	0.755	0.3617	1.4629	0.755	0.7257	0.1850	0.0888
No log	5.0	125	0.7616	0.765	0.3363	1.4885	0.765	0.7391	0.2017	0.0843
No log	6.0	150	1.0029	0.72	0.4200	1.6550	0.72	0.7047	0.2303	0.1169
No log	7.0	175	0.6286	0.825	0.2766	1.2493	0.825	0.7930	0.1954	0.0646
No log	8.0	200	0.6859	0.82	0.2857	1.4847	0.82	0.7971	0.1837	0.0699
No log	9.0	225	0.6365	0.81	0.2765	1.1457	0.81	0.7913	0.1604	0.0669
No log	10.0	250	0.6085	0.81	0.2614	1.5809	0.81	0.7928	0.1874	0.0536
No log	11.0	275	0.5900	0.84	0.2620	1.1457	0.8400	0.8308	0.1695	0.0674
No log	12.0	300	0.8544	0.75	0.3667	1.9577	0.75	0.7330	0.1988	0.1329
No log	13.0	325	0.5265	0.845	0.2278	1.2521	0.845	0.8209	0.1518	0.0399
No log	14.0	350	0.5702	0.815	0.2567	1.5233	0.815	0.8032	0.1551	0.0519
No log	15.0	375	0.5933	0.845	0.2581	1.4776	0.845	0.8341	0.1659	0.0738
No log	16.0	400	0.5697	0.84	0.2496	1.6732	0.8400	0.8235	0.1470	0.0557
No log	17.0	425	0.5471	0.825	0.2428	1.7010	0.825	0.8093	0.1406	0.0461
No log	18.0	450	0.5696	0.825	0.2546	1.4095	0.825	0.7977	0.1461	0.0612
No log	19.0	475	0.6544	0.805	0.2959	1.8251	0.805	0.7970	0.1681	0.0605
0.4416	20.0	500	0.5113	0.83	0.2327	1.4103	0.83	0.8093	0.1380	0.0541
0.4416	21.0	525	0.5255	0.84	0.2375	1.6750	0.8400	0.8220	0.1320	0.0462
0.4416	22.0	550	0.5889	0.835	0.2681	1.7850	0.835	0.8242	0.1507	0.0683
0.4416	23.0	575	0.5456	0.835	0.2492	1.8481	0.835	0.8137	0.1716	0.0550
0.4416	24.0	600	0.5661	0.83	0.2611	1.8434	0.83	0.8156	0.1618	0.0591
0.4416	25.0	625	0.5444	0.83	0.2484	1.7579	0.83	0.8091	0.1478	0.0530
0.4416	26.0	650	0.5418	0.83	0.2503	1.7188	0.83	0.8125	0.1564	0.0484
0.4416	27.0	675	0.5532	0.835	0.2540	1.8931	0.835	0.8146	0.1694	0.0514
0.4416	28.0	700	0.5492	0.835	0.2518	1.8959	0.835	0.8155	0.1505	0.0495
0.4416	29.0	725	0.5478	0.825	0.2505	1.8907	0.825	0.8069	0.1548	0.0503
0.4416	30.0	750	0.5478	0.835	0.2510	1.8881	0.835	0.8178	0.1467	0.0521
0.4416	31.0	775	0.5472	0.825	0.2505	1.8888	0.825	0.8064	0.1527	0.0510
0.4416	32.0	800	0.5522	0.83	0.2527	1.8927	0.83	0.8126	0.1449	0.0520
0.4416	33.0	825	0.5513	0.825	0.2524	1.8989	0.825	0.8064	0.1625	0.0509
0.4416	34.0	850	0.5465	0.835	0.2504	1.8880	0.835	0.8148	0.1519	0.0520
0.4416	35.0	875	0.5489	0.825	0.2515	1.8866	0.825	0.8064	0.1538	0.0510
0.4416	36.0	900	0.5508	0.825	0.2521	1.8922	0.825	0.8053	0.1356	0.0526
0.4416	37.0	925	0.5495	0.825	0.2522	1.8881	0.825	0.8064	0.1517	0.0514
0.4416	38.0	950	0.5483	0.825	0.2514	1.8859	0.825	0.8064	0.1749	0.0511
0.4416	39.0	975	0.5508	0.825	0.2524	1.8868	0.825	0.8064	0.1459	0.0514
0.0519	40.0	1000	0.5519	0.825	0.2529	1.8862	0.825	0.8064	0.1532	0.0513
0.0519	41.0	1025	0.5522	0.825	0.2530	1.8882	0.825	0.8064	0.1665	0.0519
0.0519	42.0	1050	0.5507	0.825	0.2525	1.8870	0.825	0.8064	0.1613	0.0508
0.0519	43.0	1075	0.5528	0.825	0.2536	1.8884	0.825	0.8064	0.1634	0.0517
0.0519	44.0	1100	0.5520	0.825	0.2531	1.8879	0.825	0.8064	0.1519	0.0525
0.0519	45.0	1125	0.5524	0.825	0.2535	1.8876	0.825	0.8053	0.1582	0.0515
0.0519	46.0	1150	0.5525	0.825	0.2534	1.8867	0.825	0.8064	0.1592	0.0519
0.0519	47.0	1175	0.5532	0.825	0.2539	1.8875	0.825	0.8064	0.1621	0.0521
0.0519	48.0	1200	0.5540	0.825	0.2540	1.8865	0.825	0.8064	0.1502	0.0522
0.0519	49.0	1225	0.5523	0.825	0.2538	1.8268	0.825	0.8072	0.1625	0.0514
0.0519	50.0	1250	0.5535	0.825	0.2539	1.8871	0.825	0.8064	0.1684	0.0517
0.0519	51.0	1275	0.5526	0.825	0.2534	1.8850	0.825	0.8064	0.1621	0.0519
0.0519	52.0	1300	0.5543	0.825	0.2543	1.8865	0.825	0.8064	0.1429	0.0521
0.0519	53.0	1325	0.5526	0.825	0.2538	1.8866	0.825	0.8064	0.1613	0.0515
0.0519	54.0	1350	0.5530	0.82	0.2538	1.8877	0.82	0.8009	0.1620	0.0518
0.0519	55.0	1375	0.5550	0.825	0.2547	1.8872	0.825	0.8064	0.1567	0.0522
0.0519	56.0	1400	0.5565	0.825	0.2552	1.8859	0.825	0.8064	0.1400	0.0523
0.0519	57.0	1425	0.5552	0.825	0.2548	1.8874	0.825	0.8064	0.1543	0.0520
0.0519	58.0	1450	0.5537	0.825	0.2542	1.8860	0.825	0.8064	0.1531	0.0516
0.0519	59.0	1475	0.5559	0.825	0.2551	1.8879	0.825	0.8064	0.1564	0.0525
0.0508	60.0	1500	0.5548	0.825	0.2545	1.8866	0.825	0.8064	0.1526	0.0522
0.0508	61.0	1525	0.5557	0.825	0.2550	1.8884	0.825	0.8064	0.1443	0.0524
0.0508	62.0	1550	0.5548	0.82	0.2546	1.8874	0.82	0.8009	0.1709	0.0527
0.0508	63.0	1575	0.5556	0.825	0.2551	1.8899	0.825	0.8064	0.1606	0.0524
0.0508	64.0	1600	0.5562	0.825	0.2553	1.8872	0.825	0.8064	0.1467	0.0527
0.0508	65.0	1625	0.5569	0.825	0.2554	1.8879	0.825	0.8064	0.1537	0.0524
0.0508	66.0	1650	0.5567	0.825	0.2555	1.8873	0.825	0.8064	0.1601	0.0525
0.0508	67.0	1675	0.5556	0.825	0.2550	1.8878	0.825	0.8064	0.1601	0.0527
0.0508	68.0	1700	0.5570	0.825	0.2555	1.8879	0.825	0.8064	0.1679	0.0528
0.0508	69.0	1725	0.5560	0.825	0.2553	1.8886	0.825	0.8064	0.1525	0.0521
0.0508	70.0	1750	0.5562	0.825	0.2553	1.8878	0.825	0.8064	0.1531	0.0528
0.0508	71.0	1775	0.5572	0.82	0.2557	1.8883	0.82	0.8009	0.1718	0.0530
0.0508	72.0	1800	0.5567	0.82	0.2555	1.8888	0.82	0.8009	0.1630	0.0525
0.0508	73.0	1825	0.5571	0.825	0.2556	1.8882	0.825	0.8064	0.1598	0.0528
0.0508	74.0	1850	0.5580	0.825	0.2561	1.8901	0.825	0.8064	0.1543	0.0530
0.0508	75.0	1875	0.5579	0.82	0.2561	1.8892	0.82	0.8009	0.1721	0.0530
0.0508	76.0	1900	0.5574	0.82	0.2559	1.8892	0.82	0.8009	0.1636	0.0528
0.0508	77.0	1925	0.5569	0.82	0.2557	1.8393	0.82	0.8009	0.1634	0.0526
0.0508	78.0	1950	0.5572	0.82	0.2558	1.8887	0.82	0.8009	0.1637	0.0530
0.0508	79.0	1975	0.5578	0.82	0.2560	1.8888	0.82	0.8009	0.1579	0.0530
0.0507	80.0	2000	0.5577	0.82	0.2559	1.8889	0.82	0.8009	0.1578	0.0532
0.0507	81.0	2025	0.5578	0.82	0.2560	1.8889	0.82	0.8009	0.1578	0.0533
0.0507	82.0	2050	0.5579	0.825	0.2561	1.8891	0.825	0.8064	0.1602	0.0528
0.0507	83.0	2075	0.5581	0.825	0.2562	1.8894	0.825	0.8064	0.1544	0.0528
0.0507	84.0	2100	0.5579	0.82	0.2561	1.8894	0.82	0.8009	0.1581	0.0531
0.0507	85.0	2125	0.5580	0.82	0.2561	1.8896	0.82	0.8009	0.1578	0.0528
0.0507	86.0	2150	0.5581	0.82	0.2562	1.8891	0.82	0.8009	0.1580	0.0532
0.0507	87.0	2175	0.5582	0.82	0.2562	1.8467	0.82	0.8009	0.1581	0.0528
0.0507	88.0	2200	0.5583	0.82	0.2562	1.8891	0.82	0.8009	0.1580	0.0531
0.0507	89.0	2225	0.5584	0.815	0.2563	1.8894	0.815	0.7976	0.1608	0.0534
0.0507	90.0	2250	0.5578	0.82	0.2561	1.8894	0.82	0.8009	0.1578	0.0530
0.0507	91.0	2275	0.5584	0.815	0.2563	1.8896	0.815	0.7976	0.1607	0.0532
0.0507	92.0	2300	0.5583	0.82	0.2562	1.8893	0.82	0.8009	0.1581	0.0531
0.0507	93.0	2325	0.5582	0.82	0.2562	1.8898	0.82	0.8009	0.1579	0.0530
0.0507	94.0	2350	0.5582	0.82	0.2562	1.8392	0.82	0.8009	0.1578	0.0530
0.0507	95.0	2375	0.5584	0.82	0.2563	1.8897	0.82	0.8009	0.1582	0.0531
0.0507	96.0	2400	0.5582	0.82	0.2562	1.8898	0.82	0.8009	0.1578	0.0530
0.0507	97.0	2425	0.5583	0.82	0.2563	1.8896	0.82	0.8009	0.1580	0.0530
0.0507	98.0	2450	0.5582	0.82	0.2562	1.8898	0.82	0.8009	0.1578	0.0530
0.0507	99.0	2475	0.5583	0.82	0.2563	1.8898	0.82	0.8009	0.1578	0.0530
0.0507	100.0	2500	0.5583	0.82	0.2563	1.8898	0.82	0.8009	0.1578	0.0530

Framework versions

Transformers 4.30.2
Pytorch 1.13.1
Datasets 2.13.1
Tokenizers 0.13.3

ayanban011/6_e_200-tiny_tobacco3482_kd_CEKD_t5.0_a0.9

6_e_200-tiny_tobacco3482_kd_CEKD_t5.0_a0.9

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions