swin-base_tobacco

This model is a fine-tuned version of microsoft/swinv2-base-patch4-window8-256 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6059
• Accuracy: 0.835
• Brier Loss: 0.2576
• Nll: 1.2824
• F1 Micro: 0.835
• F1 Macro: 0.8348
• Ece: 0.1310
• Aurc: 0.0387

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 256
• 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	0.96	3	2.3165	0.11	0.9031	7.6310	0.11	0.0604	0.2004	0.8718
No log	1.96	6	2.2894	0.155	0.8975	6.8146	0.155	0.0944	0.2230	0.8555
No log	2.96	9	2.2481	0.215	0.8888	5.1480	0.2150	0.1472	0.2492	0.8119
No log	3.96	12	2.1955	0.275	0.8770	4.2879	0.275	0.1939	0.2844	0.6562
No log	4.96	15	2.1326	0.36	0.8619	3.8809	0.36	0.2199	0.3357	0.4962
No log	5.96	18	2.0568	0.375	0.8415	3.9254	0.375	0.2309	0.3377	0.4471
No log	6.96	21	1.9639	0.375	0.8126	3.8158	0.375	0.2319	0.3195	0.4534
No log	7.96	24	1.8621	0.375	0.7781	3.3244	0.375	0.2456	0.2924	0.4833
No log	8.96	27	1.7100	0.44	0.7273	2.8211	0.44	0.3136	0.3188	0.3515
No log	9.96	30	1.5377	0.535	0.6611	2.4560	0.535	0.4259	0.3557	0.2259
No log	10.96	33	1.3588	0.595	0.5825	2.3216	0.595	0.4933	0.2986	0.1795
No log	11.96	36	1.2072	0.62	0.5215	2.3831	0.62	0.5352	0.2927	0.1541
No log	12.96	39	1.0766	0.67	0.4715	2.2078	0.67	0.5966	0.2727	0.1219
No log	13.96	42	0.9699	0.675	0.4408	1.8028	0.675	0.5961	0.2568	0.1215
No log	14.96	45	0.8660	0.68	0.4011	1.4772	0.68	0.5978	0.2176	0.1014
No log	15.96	48	0.7907	0.725	0.3709	1.4755	0.7250	0.6768	0.2055	0.0904
No log	16.96	51	0.7362	0.75	0.3501	1.3822	0.75	0.7077	0.2042	0.0806
No log	17.96	54	0.6867	0.76	0.3322	1.3191	0.76	0.7177	0.1926	0.0724
No log	18.96	57	0.6572	0.78	0.3203	1.2996	0.78	0.7424	0.1920	0.0699
No log	19.96	60	0.6074	0.785	0.2967	1.3136	0.785	0.7686	0.1705	0.0589
No log	20.96	63	0.6050	0.795	0.2956	1.3729	0.795	0.7793	0.1762	0.0600
No log	21.96	66	0.5748	0.83	0.2785	1.3558	0.83	0.8113	0.1744	0.0529
No log	22.96	69	0.5722	0.815	0.2756	1.3937	0.815	0.8097	0.1767	0.0489
No log	23.96	72	0.5689	0.795	0.2750	1.3641	0.795	0.7947	0.1452	0.0539
No log	24.96	75	0.5536	0.825	0.2718	1.2773	0.825	0.8068	0.1698	0.0509
No log	25.96	78	0.5464	0.805	0.2726	1.2772	0.805	0.7888	0.1499	0.0487
No log	26.96	81	0.5455	0.81	0.2626	1.3607	0.81	0.8080	0.1750	0.0471
No log	27.96	84	0.5542	0.815	0.2609	1.3643	0.815	0.8089	0.1521	0.0466
No log	28.96	87	0.5480	0.82	0.2710	1.2996	0.82	0.8227	0.1422	0.0468
No log	29.96	90	0.5507	0.83	0.2654	1.3425	0.83	0.8320	0.1491	0.0475
No log	30.96	93	0.5608	0.815	0.2591	1.4365	0.815	0.8145	0.1405	0.0442
No log	31.96	96	0.5473	0.825	0.2622	1.3600	0.825	0.8198	0.1339	0.0424
No log	32.96	99	0.5296	0.83	0.2588	1.2906	0.83	0.8311	0.1373	0.0416
No log	33.96	102	0.5370	0.82	0.2522	1.2895	0.82	0.8214	0.1428	0.0436
No log	34.96	105	0.5578	0.8	0.2707	1.3364	0.8000	0.8056	0.1708	0.0481
No log	35.96	108	0.5193	0.825	0.2484	1.2883	0.825	0.8250	0.1316	0.0405
No log	36.96	111	0.5306	0.815	0.2569	1.2856	0.815	0.8093	0.1344	0.0420
No log	37.96	114	0.5824	0.815	0.2729	1.3994	0.815	0.8182	0.1418	0.0479
No log	38.96	117	0.5486	0.82	0.2549	1.2974	0.82	0.8259	0.1312	0.0443
No log	39.96	120	0.5421	0.83	0.2545	1.3575	0.83	0.8316	0.1491	0.0415
No log	40.96	123	0.5477	0.81	0.2700	1.3251	0.81	0.8166	0.1499	0.0418
No log	41.96	126	0.5404	0.825	0.2553	1.3186	0.825	0.8309	0.1519	0.0414
No log	42.96	129	0.5698	0.83	0.2598	1.3249	0.83	0.8386	0.1396	0.0452
No log	43.96	132	0.5538	0.815	0.2605	1.3122	0.815	0.8212	0.1410	0.0430
No log	44.96	135	0.5369	0.81	0.2586	1.3030	0.81	0.8141	0.1404	0.0409
No log	45.96	138	0.5614	0.825	0.2615	1.3881	0.825	0.8278	0.1404	0.0427
No log	46.96	141	0.5636	0.825	0.2601	1.4077	0.825	0.8286	0.1345	0.0421
No log	47.96	144	0.5783	0.83	0.2684	1.3350	0.83	0.8304	0.1373	0.0422
No log	48.96	147	0.5749	0.825	0.2663	1.3167	0.825	0.8241	0.1308	0.0424
No log	49.96	150	0.5802	0.82	0.2692	1.3191	0.82	0.8194	0.1217	0.0461
No log	50.96	153	0.5696	0.82	0.2639	1.3330	0.82	0.8175	0.1372	0.0429
No log	51.96	156	0.5827	0.84	0.2656	1.3975	0.8400	0.8444	0.1378	0.0426
No log	52.96	159	0.5725	0.805	0.2669	1.3172	0.805	0.7997	0.1459	0.0422
No log	53.96	162	0.5769	0.805	0.2691	1.3111	0.805	0.7991	0.1457	0.0434
No log	54.96	165	0.5883	0.805	0.2647	1.4581	0.805	0.8104	0.1405	0.0430
No log	55.96	168	0.5834	0.835	0.2543	1.4586	0.835	0.8349	0.1346	0.0407
No log	56.96	171	0.5875	0.835	0.2543	1.3211	0.835	0.8358	0.1320	0.0402
No log	57.96	174	0.5741	0.84	0.2533	1.3027	0.8400	0.8405	0.1290	0.0395
No log	58.96	177	0.5737	0.82	0.2624	1.3104	0.82	0.8167	0.1437	0.0396
No log	59.96	180	0.5796	0.815	0.2603	1.4021	0.815	0.8154	0.1286	0.0406
No log	60.96	183	0.5711	0.83	0.2553	1.4016	0.83	0.8306	0.1272	0.0390
No log	61.96	186	0.5670	0.825	0.2591	1.3136	0.825	0.8263	0.1429	0.0406
No log	62.96	189	0.5736	0.825	0.2592	1.3077	0.825	0.8231	0.1244	0.0417
No log	63.96	192	0.5730	0.83	0.2531	1.3007	0.83	0.8274	0.1275	0.0401
No log	64.96	195	0.6130	0.82	0.2687	1.3014	0.82	0.8246	0.1484	0.0414
No log	65.96	198	0.6023	0.825	0.2596	1.3107	0.825	0.8254	0.1373	0.0404
No log	66.96	201	0.5923	0.825	0.2599	1.3078	0.825	0.8263	0.1312	0.0411
No log	67.96	204	0.6197	0.81	0.2766	1.3046	0.81	0.8035	0.1373	0.0451
No log	68.96	207	0.5918	0.805	0.2651	1.3019	0.805	0.8044	0.1407	0.0404
No log	69.96	210	0.5908	0.835	0.2544	1.3286	0.835	0.8344	0.1354	0.0394
No log	70.96	213	0.5941	0.83	0.2558	1.3019	0.83	0.8324	0.1402	0.0401
No log	71.96	216	0.5994	0.82	0.2588	1.2998	0.82	0.8215	0.1297	0.0411
No log	72.96	219	0.6083	0.825	0.2638	1.3525	0.825	0.8257	0.1379	0.0410
No log	73.96	222	0.5980	0.825	0.2609	1.3515	0.825	0.8295	0.1457	0.0394
No log	74.96	225	0.5945	0.83	0.2568	1.3670	0.83	0.8302	0.1324	0.0390
No log	75.96	228	0.5982	0.845	0.2535	1.4552	0.845	0.8476	0.1246	0.0390
No log	76.96	231	0.5850	0.83	0.2507	1.3700	0.83	0.8287	0.1348	0.0391
No log	77.96	234	0.5859	0.825	0.2566	1.2917	0.825	0.8232	0.1309	0.0394
No log	78.96	237	0.6085	0.835	0.2630	1.3516	0.835	0.8370	0.1329	0.0420
No log	79.96	240	0.6108	0.835	0.2621	1.2943	0.835	0.8370	0.1395	0.0414
No log	80.96	243	0.6061	0.81	0.2596	1.2898	0.81	0.8119	0.1313	0.0413
No log	81.96	246	0.6006	0.815	0.2564	1.2952	0.815	0.8122	0.1453	0.0406
No log	82.96	249	0.6050	0.825	0.2577	1.2998	0.825	0.8283	0.1271	0.0400
No log	83.96	252	0.6197	0.835	0.2658	1.3021	0.835	0.8386	0.1222	0.0414
No log	84.96	255	0.6086	0.825	0.2651	1.2889	0.825	0.8251	0.1207	0.0404
No log	85.96	258	0.5965	0.83	0.2587	1.2929	0.83	0.8304	0.1323	0.0397
No log	86.96	261	0.5897	0.82	0.2550	1.2980	0.82	0.8171	0.1372	0.0394
No log	87.96	264	0.5887	0.83	0.2551	1.2950	0.83	0.8290	0.1251	0.0391
No log	88.96	267	0.5958	0.82	0.2598	1.2871	0.82	0.8180	0.1319	0.0392
No log	89.96	270	0.6088	0.82	0.2658	1.2805	0.82	0.8184	0.1513	0.0396
No log	90.96	273	0.6192	0.825	0.2692	1.2772	0.825	0.8263	0.1258	0.0402
No log	91.96	276	0.6230	0.825	0.2689	1.2777	0.825	0.8263	0.1416	0.0404
No log	92.96	279	0.6223	0.83	0.2667	1.2792	0.83	0.8318	0.1296	0.0401
No log	93.96	282	0.6145	0.83	0.2627	1.2797	0.83	0.8321	0.1265	0.0394
No log	94.96	285	0.6105	0.83	0.2610	1.2807	0.83	0.8321	0.1352	0.0392
No log	95.96	288	0.6095	0.83	0.2602	1.2815	0.83	0.8321	0.1360	0.0390
No log	96.96	291	0.6076	0.835	0.2590	1.2824	0.835	0.8348	0.1255	0.0389
No log	97.96	294	0.6060	0.835	0.2578	1.2827	0.835	0.8348	0.1281	0.0388
No log	98.96	297	0.6058	0.835	0.2575	1.2825	0.835	0.8348	0.1410	0.0387
No log	99.96	300	0.6059	0.835	0.2576	1.2824	0.835	0.8348	0.1310	0.0387

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1.post200
• Datasets 2.9.0
• Tokenizers 0.13.2