dit-base_tobacco-small_tobacco3482_simkd

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: 0.6962
Accuracy: 0.85
Brier Loss: 0.2700
Nll: 0.9667
F1 Micro: 0.85
F1 Macro: 0.8241
Ece: 0.2479
Aurc: 0.0379

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: 16
eval_batch_size: 16
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	50	1.0013	0.18	0.8965	4.5407	0.18	0.1379	0.2160	0.6680
No log	2.0	100	0.9916	0.3	0.8871	3.1090	0.3	0.1526	0.3057	0.4735
No log	3.0	150	0.9644	0.51	0.8433	2.4502	0.51	0.3257	0.4499	0.2544
No log	4.0	200	0.9207	0.575	0.7585	2.1964	0.575	0.3958	0.4563	0.2193
No log	5.0	250	0.8726	0.635	0.6620	2.3923	0.635	0.5105	0.4321	0.1730
No log	6.0	300	0.8303	0.665	0.5604	1.4922	0.665	0.5869	0.3717	0.1305
No log	7.0	350	0.7994	0.745	0.4490	1.3772	0.745	0.6541	0.3557	0.0853
No log	8.0	400	0.7822	0.79	0.4124	1.2076	0.79	0.7109	0.3035	0.0873
No log	9.0	450	0.7808	0.78	0.3955	1.5529	0.78	0.7041	0.3123	0.0763
0.8704	10.0	500	0.7923	0.695	0.4296	1.7171	0.695	0.6150	0.3012	0.1039
0.8704	11.0	550	0.7848	0.745	0.4327	1.6327	0.745	0.6972	0.2800	0.1321
0.8704	12.0	600	0.7600	0.825	0.3579	1.2569	0.825	0.7621	0.3015	0.0624
0.8704	13.0	650	0.7570	0.79	0.3554	1.4638	0.79	0.7706	0.2964	0.0621
0.8704	14.0	700	0.7504	0.81	0.3434	1.5597	0.81	0.7714	0.2930	0.0589
0.8704	15.0	750	0.7481	0.8	0.3439	1.3827	0.8000	0.7641	0.2805	0.0675
0.8704	16.0	800	0.7358	0.81	0.3357	1.4522	0.81	0.7889	0.3077	0.0610
0.8704	17.0	850	0.7294	0.82	0.3179	1.0458	0.82	0.7820	0.2909	0.0564
0.8704	18.0	900	0.7229	0.815	0.3092	1.2562	0.815	0.7862	0.2719	0.0496
0.8704	19.0	950	0.7186	0.825	0.3069	1.0425	0.825	0.7977	0.2824	0.0558
0.6968	20.0	1000	0.7156	0.83	0.3031	0.9897	0.83	0.8039	0.2660	0.0490
0.6968	21.0	1050	0.7135	0.82	0.3014	1.0562	0.82	0.7887	0.2745	0.0462
0.6968	22.0	1100	0.7116	0.835	0.2997	0.9822	0.835	0.8102	0.2817	0.0452
0.6968	23.0	1150	0.7114	0.82	0.3047	0.9197	0.82	0.7937	0.2669	0.0484
0.6968	24.0	1200	0.7111	0.8	0.3032	0.9744	0.8000	0.7690	0.2624	0.0504
0.6968	25.0	1250	0.7076	0.805	0.3025	0.9884	0.805	0.7677	0.2538	0.0478
0.6968	26.0	1300	0.7074	0.82	0.3037	0.9954	0.82	0.7877	0.2592	0.0496
0.6968	27.0	1350	0.7053	0.825	0.2998	0.9712	0.825	0.7885	0.2628	0.0454
0.6968	28.0	1400	0.7046	0.82	0.2936	0.9780	0.82	0.7886	0.2573	0.0438
0.6968	29.0	1450	0.7068	0.82	0.3000	0.9943	0.82	0.7895	0.2382	0.0447
0.6551	30.0	1500	0.7045	0.83	0.2881	0.9107	0.83	0.8010	0.2363	0.0439
0.6551	31.0	1550	0.7033	0.825	0.2936	0.9794	0.825	0.7858	0.2556	0.0433
0.6551	32.0	1600	0.7014	0.82	0.2890	0.9799	0.82	0.7895	0.2495	0.0418
0.6551	33.0	1650	0.7020	0.815	0.2921	0.9658	0.815	0.7820	0.2556	0.0449
0.6551	34.0	1700	0.7012	0.835	0.2885	1.0419	0.835	0.8042	0.2581	0.0417
0.6551	35.0	1750	0.7013	0.835	0.2902	0.9773	0.835	0.8035	0.2522	0.0435
0.6551	36.0	1800	0.7016	0.825	0.2884	0.9815	0.825	0.7851	0.2518	0.0432
0.6551	37.0	1850	0.7007	0.835	0.2888	0.9724	0.835	0.8133	0.2486	0.0438
0.6551	38.0	1900	0.6984	0.825	0.2847	0.9650	0.825	0.7897	0.2487	0.0415
0.6551	39.0	1950	0.7001	0.84	0.2843	1.0535	0.8400	0.8104	0.2566	0.0418
0.6381	40.0	2000	0.6990	0.825	0.2843	0.9673	0.825	0.7963	0.2396	0.0429
0.6381	41.0	2050	0.7002	0.84	0.2875	1.0599	0.8400	0.8098	0.2618	0.0413
0.6381	42.0	2100	0.6967	0.83	0.2791	0.9676	0.83	0.7929	0.2441	0.0403
0.6381	43.0	2150	0.6978	0.835	0.2802	0.9771	0.835	0.8071	0.2526	0.0416
0.6381	44.0	2200	0.6969	0.84	0.2795	0.9478	0.8400	0.8164	0.2464	0.0418
0.6381	45.0	2250	0.6971	0.835	0.2760	0.9712	0.835	0.8030	0.2333	0.0392
0.6381	46.0	2300	0.6985	0.84	0.2813	0.9692	0.8400	0.8072	0.2403	0.0404
0.6381	47.0	2350	0.6976	0.835	0.2796	1.0420	0.835	0.8042	0.2374	0.0406
0.6381	48.0	2400	0.6965	0.85	0.2778	0.9753	0.85	0.8205	0.2653	0.0403
0.6381	49.0	2450	0.6969	0.825	0.2747	0.9606	0.825	0.7871	0.2478	0.0394
0.6274	50.0	2500	0.6954	0.835	0.2746	0.9572	0.835	0.8070	0.2395	0.0406
0.6274	51.0	2550	0.6972	0.835	0.2755	1.0383	0.835	0.8070	0.2484	0.0391
0.6274	52.0	2600	0.6955	0.83	0.2752	0.9699	0.83	0.7998	0.2562	0.0406
0.6274	53.0	2650	0.6950	0.835	0.2693	0.9563	0.835	0.8030	0.2300	0.0373
0.6274	54.0	2700	0.6960	0.83	0.2727	0.9646	0.83	0.7977	0.2347	0.0399
0.6274	55.0	2750	0.6946	0.83	0.2711	0.9603	0.83	0.8058	0.2279	0.0384
0.6274	56.0	2800	0.6940	0.835	0.2726	0.9579	0.835	0.8088	0.2478	0.0380
0.6274	57.0	2850	0.6951	0.835	0.2732	0.9594	0.835	0.8090	0.2336	0.0418
0.6274	58.0	2900	0.6936	0.84	0.2684	0.9575	0.8400	0.8079	0.2490	0.0373
0.6274	59.0	2950	0.6949	0.835	0.2701	0.9543	0.835	0.8088	0.2261	0.0389
0.6207	60.0	3000	0.6939	0.84	0.2697	0.9574	0.8400	0.8161	0.2339	0.0378
0.6207	61.0	3050	0.6952	0.84	0.2706	0.9611	0.8400	0.8080	0.2306	0.0379
0.6207	62.0	3100	0.6940	0.835	0.2691	0.9523	0.835	0.8086	0.2451	0.0382
0.6207	63.0	3150	0.6946	0.835	0.2672	0.9627	0.835	0.8088	0.2347	0.0374
0.6207	64.0	3200	0.6949	0.84	0.2713	0.9602	0.8400	0.8139	0.2404	0.0384
0.6207	65.0	3250	0.6944	0.835	0.2662	0.9603	0.835	0.8079	0.2308	0.0377
0.6207	66.0	3300	0.6946	0.835	0.2698	0.9593	0.835	0.8088	0.2352	0.0390
0.6207	67.0	3350	0.6934	0.83	0.2658	0.9558	0.83	0.8060	0.2260	0.0384
0.6207	68.0	3400	0.6944	0.83	0.2689	0.9517	0.83	0.8058	0.2208	0.0399
0.6207	69.0	3450	0.6946	0.835	0.2698	0.9553	0.835	0.8042	0.2331	0.0383
0.6156	70.0	3500	0.6948	0.83	0.2690	0.9549	0.83	0.8058	0.2280	0.0391
0.6156	71.0	3550	0.6936	0.84	0.2676	0.9532	0.8400	0.8122	0.2346	0.0383
0.6156	72.0	3600	0.6946	0.835	0.2667	0.9545	0.835	0.8088	0.2492	0.0379
0.6156	73.0	3650	0.6939	0.84	0.2670	0.9534	0.8400	0.8139	0.2466	0.0377
0.6156	74.0	3700	0.6948	0.835	0.2695	0.9522	0.835	0.8086	0.2312	0.0390
0.6156	75.0	3750	0.6951	0.835	0.2701	0.9622	0.835	0.8111	0.2158	0.0397
0.6156	76.0	3800	0.6949	0.84	0.2682	0.9606	0.8400	0.8139	0.2415	0.0382
0.6156	77.0	3850	0.6950	0.84	0.2684	0.9629	0.8400	0.8118	0.2493	0.0381
0.6156	78.0	3900	0.6946	0.835	0.2685	0.9522	0.835	0.8111	0.2360	0.0390
0.6156	79.0	3950	0.6944	0.84	0.2668	0.9544	0.8400	0.8118	0.2377	0.0372
0.612	80.0	4000	0.6954	0.84	0.2692	0.9579	0.8400	0.8139	0.2321	0.0381
0.612	81.0	4050	0.6956	0.84	0.2701	0.9606	0.8400	0.8139	0.2354	0.0382
0.612	82.0	4100	0.6952	0.835	0.2686	0.9600	0.835	0.8086	0.2540	0.0381
0.612	83.0	4150	0.6955	0.835	0.2689	0.9571	0.835	0.8086	0.2465	0.0383
0.612	84.0	4200	0.6952	0.84	0.2689	0.9583	0.8400	0.8159	0.2308	0.0387
0.612	85.0	4250	0.6956	0.835	0.2702	0.9618	0.835	0.8042	0.2365	0.0386
0.612	86.0	4300	0.6950	0.835	0.2683	0.9572	0.835	0.8086	0.2228	0.0382
0.612	87.0	4350	0.6949	0.84	0.2692	0.9583	0.8400	0.8118	0.2497	0.0381
0.612	88.0	4400	0.6953	0.845	0.2695	0.9617	0.845	0.8209	0.2558	0.0386
0.612	89.0	4450	0.6952	0.845	0.2689	0.9611	0.845	0.8209	0.2251	0.0383
0.6097	90.0	4500	0.6961	0.835	0.2701	0.9645	0.835	0.8042	0.2444	0.0386
0.6097	91.0	4550	0.6954	0.845	0.2689	0.9619	0.845	0.8209	0.2324	0.0383
0.6097	92.0	4600	0.6959	0.845	0.2700	0.9636	0.845	0.8209	0.2277	0.0388
0.6097	93.0	4650	0.6959	0.85	0.2694	0.9654	0.85	0.8241	0.2396	0.0379
0.6097	94.0	4700	0.6960	0.85	0.2696	0.9643	0.85	0.8241	0.2471	0.0379
0.6097	95.0	4750	0.6959	0.85	0.2694	0.9650	0.85	0.8241	0.2233	0.0378
0.6097	96.0	4800	0.6962	0.845	0.2700	0.9666	0.845	0.8144	0.2558	0.0382
0.6097	97.0	4850	0.6962	0.85	0.2699	0.9662	0.85	0.8241	0.2400	0.0381
0.6097	98.0	4900	0.6962	0.85	0.2700	0.9662	0.85	0.8241	0.2396	0.0380
0.6097	99.0	4950	0.6963	0.85	0.2700	0.9667	0.85	0.8241	0.2478	0.0379
0.6083	100.0	5000	0.6962	0.85	0.2700	0.9667	0.85	0.8241	0.2479	0.0379

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_simkd

dit-base_tobacco-small_tobacco3482_simkd

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions