resnet101_rvl-cdip-cnn_rvl_cdip-NK1000_kd_rand

This model is a fine-tuned version of bdpc/resnet101_rvl-cdip on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7141
• Accuracy: 0.776
• Brier Loss: 0.3162
• Nll: 1.9406
• F1 Micro: 0.776
• F1 Macro: 0.7755
• Ece: 0.0376
• Aurc: 0.0655

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: 64
• eval_batch_size: 64
• 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Brier Loss	Nll	F1 Micro	F1 Macro	Ece	Aurc
No log	1.0	250	3.5746	0.2293	0.8509	3.6359	0.2293	0.1928	0.0536	0.5849
3.6362	2.0	500	3.4699	0.2848	0.8495	4.0612	0.2848	0.2468	0.0791	0.5541
3.6362	3.0	750	2.9284	0.3957	0.7832	3.1766	0.3957	0.3732	0.1412	0.4345
2.5616	4.0	1000	2.4586	0.4497	0.7009	3.0303	0.4497	0.4362	0.0977	0.3388
2.5616	5.0	1250	2.5238	0.4268	0.7479	3.0431	0.4268	0.4163	0.2085	0.3343
1.8525	6.0	1500	1.6707	0.5915	0.5475	2.6137	0.5915	0.5880	0.0929	0.1837
1.8525	7.0	1750	1.3529	0.6385	0.4901	2.3981	0.6385	0.6360	0.0936	0.1504
1.2522	8.0	2000	1.2537	0.6585	0.4640	2.2970	0.6585	0.6591	0.0722	0.1341
1.2522	9.0	2250	1.0462	0.7053	0.4096	2.2117	0.7053	0.7032	0.0668	0.1048
0.8788	10.0	2500	1.2045	0.6743	0.4500	2.2939	0.6743	0.6751	0.0818	0.1245
0.8788	11.0	2750	1.0487	0.7045	0.4107	2.1999	0.7045	0.6995	0.0686	0.1074
0.6325	12.0	3000	1.0625	0.7105	0.4075	2.2241	0.7105	0.7038	0.0670	0.1062
0.6325	13.0	3250	0.9567	0.7262	0.3820	2.1370	0.7262	0.7211	0.0585	0.0950
0.4724	14.0	3500	0.8833	0.744	0.3598	2.0806	0.744	0.7450	0.0477	0.0820
0.4724	15.0	3750	0.9074	0.744	0.3638	2.1065	0.744	0.7447	0.0504	0.0869
0.3827	16.0	4000	0.9355	0.7382	0.3678	2.1484	0.7382	0.7369	0.0490	0.0877
0.3827	17.0	4250	0.8586	0.7502	0.3508	2.0839	0.7502	0.7495	0.0500	0.0803
0.3253	18.0	4500	0.8572	0.7412	0.3579	2.0452	0.7412	0.7419	0.0504	0.0830
0.3253	19.0	4750	0.8663	0.741	0.3581	2.0923	0.7410	0.7405	0.0481	0.0826
0.288	20.0	5000	0.8162	0.7535	0.3431	2.0485	0.7535	0.7547	0.0444	0.0776
0.288	21.0	5250	0.9363	0.735	0.3739	2.1240	0.735	0.7332	0.0482	0.0895
0.2614	22.0	5500	0.8013	0.762	0.3391	2.0235	0.762	0.7598	0.0457	0.0743
0.2614	23.0	5750	0.7936	0.756	0.3391	2.0319	0.756	0.7585	0.0450	0.0748
0.2434	24.0	6000	0.8142	0.7515	0.3476	1.9971	0.7515	0.7483	0.0455	0.0769
0.2434	25.0	6250	0.8032	0.757	0.3441	2.0560	0.757	0.7569	0.0482	0.0766
0.2252	26.0	6500	0.7746	0.7622	0.3305	2.0183	0.7622	0.7648	0.0412	0.0717
0.2252	27.0	6750	0.7663	0.7655	0.3342	1.9947	0.7655	0.7674	0.0371	0.0724
0.213	28.0	7000	0.8283	0.756	0.3460	2.0260	0.756	0.7558	0.0432	0.0770
0.213	29.0	7250	0.7564	0.7638	0.3301	2.0031	0.7638	0.7653	0.0402	0.0716
0.2007	30.0	7500	0.7836	0.761	0.3376	2.0263	0.761	0.7599	0.0380	0.0741
0.2007	31.0	7750	0.7539	0.7668	0.3297	1.9693	0.7668	0.7647	0.0346	0.0699
0.1911	32.0	8000	0.7534	0.7665	0.3294	1.9842	0.7665	0.7645	0.0449	0.0702
0.1911	33.0	8250	0.7608	0.769	0.3276	2.0015	0.769	0.7687	0.0381	0.0701
0.183	34.0	8500	0.7511	0.7675	0.3273	1.9691	0.7675	0.7666	0.0394	0.0687
0.183	35.0	8750	0.7582	0.7615	0.3282	1.9961	0.7615	0.7607	0.0468	0.0717
0.1749	36.0	9000	0.7272	0.7702	0.3228	1.9832	0.7702	0.7706	0.0385	0.0679
0.1749	37.0	9250	0.7446	0.7685	0.3258	2.0006	0.7685	0.7698	0.0391	0.0694
0.1672	38.0	9500	0.7252	0.773	0.3200	1.9722	0.7730	0.7719	0.0423	0.0673
0.1672	39.0	9750	0.7260	0.7738	0.3184	1.9599	0.7738	0.7741	0.0357	0.0660
0.1608	40.0	10000	0.7317	0.7715	0.3191	1.9651	0.7715	0.7717	0.0340	0.0668
0.1608	41.0	10250	0.7268	0.7742	0.3206	1.9666	0.7742	0.7728	0.0429	0.0671
0.1549	42.0	10500	0.7296	0.7745	0.3198	1.9628	0.7745	0.7738	0.0341	0.0666
0.1549	43.0	10750	0.7223	0.773	0.3173	1.9566	0.7730	0.7732	0.0414	0.0661
0.15	44.0	11000	0.7184	0.775	0.3175	1.9566	0.775	0.7746	0.0385	0.0661
0.15	45.0	11250	0.7205	0.7755	0.3181	1.9439	0.7755	0.7749	0.0377	0.0661
0.1458	46.0	11500	0.7174	0.777	0.3173	1.9560	0.777	0.7768	0.0424	0.0660
0.1458	47.0	11750	0.7152	0.7755	0.3164	1.9527	0.7755	0.7750	0.0404	0.0656
0.1429	48.0	12000	0.7173	0.777	0.3161	1.9350	0.777	0.7774	0.0415	0.0656
0.1429	49.0	12250	0.7142	0.7755	0.3159	1.9515	0.7755	0.7748	0.0404	0.0656
0.141	50.0	12500	0.7141	0.776	0.3162	1.9406	0.776	0.7755	0.0376	0.0655

Framework versions

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