resnet101_rvl-cdip-cnn_rvl_cdip-NK1000_kd_NKD_t1.0_g1.5_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: 3.0688
• Accuracy: 0.768
• Brier Loss: 0.3296
• Nll: 1.9490
• F1 Micro: 0.768
• F1 Macro: 0.7688
• Ece: 0.0595
• Aurc: 0.0737

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	5.5798	0.1893	0.8757	3.8790	0.1893	0.1519	0.0656	0.6360
5.6792	2.0	500	5.5256	0.2642	0.8415	4.3054	0.2642	0.2235	0.0705	0.5478
5.6792	3.0	750	5.3875	0.3157	0.8213	3.4839	0.3157	0.2920	0.0835	0.5208
4.8939	4.0	1000	4.4870	0.473	0.6754	2.6509	0.473	0.4671	0.0728	0.3199
4.8939	5.0	1250	4.4320	0.5015	0.6345	2.7717	0.5015	0.4779	0.0577	0.2773
4.3006	6.0	1500	4.2399	0.5435	0.5938	2.6489	0.5435	0.5334	0.0527	0.2353
4.3006	7.0	1750	3.8189	0.6155	0.5111	2.4587	0.6155	0.6096	0.0725	0.1646
3.6984	8.0	2000	3.5238	0.6795	0.4476	2.2085	0.6795	0.6810	0.0712	0.1258
3.6984	9.0	2250	3.4654	0.6877	0.4287	2.2234	0.6877	0.6852	0.0511	0.1198
3.32	10.0	2500	3.4769	0.692	0.4253	2.2586	0.692	0.6868	0.0456	0.1164
3.32	11.0	2750	3.3473	0.7235	0.3967	2.1360	0.7235	0.7245	0.0518	0.1063
3.0488	12.0	3000	3.3891	0.712	0.3929	2.1461	0.712	0.7100	0.0491	0.1043
3.0488	13.0	3250	3.3123	0.7208	0.3846	2.1236	0.7208	0.7171	0.0444	0.0958
2.8727	14.0	3500	3.3357	0.7147	0.3877	2.0928	0.7147	0.7170	0.0489	0.1000
2.8727	15.0	3750	3.2576	0.7318	0.3703	2.1052	0.7318	0.7358	0.0473	0.0901
2.7552	16.0	4000	3.2528	0.738	0.3650	2.0968	0.738	0.7396	0.0477	0.0877
2.7552	17.0	4250	3.4241	0.7093	0.4120	2.1652	0.7093	0.7078	0.0540	0.1114
2.6718	18.0	4500	3.2877	0.7358	0.3686	2.0587	0.7358	0.7362	0.0517	0.0899
2.6718	19.0	4750	3.2690	0.7375	0.3650	2.0509	0.7375	0.7409	0.0527	0.0891
2.6158	20.0	5000	3.2650	0.7398	0.3629	2.0803	0.7398	0.7432	0.0491	0.0889
2.6158	21.0	5250	3.2507	0.7452	0.3599	2.0773	0.7452	0.7467	0.0577	0.0859
2.5766	22.0	5500	3.1737	0.7525	0.3481	2.0408	0.7525	0.7539	0.0557	0.0815
2.5766	23.0	5750	3.1625	0.7585	0.3478	2.0313	0.7585	0.7599	0.0587	0.0813
2.5388	24.0	6000	3.2357	0.746	0.3563	2.0384	0.746	0.7445	0.0592	0.0838
2.5388	25.0	6250	3.1653	0.761	0.3441	2.0389	0.761	0.7605	0.0423	0.0783
2.5129	26.0	6500	3.1662	0.7532	0.3468	2.0033	0.7532	0.7556	0.0594	0.0805
2.5129	27.0	6750	3.1224	0.7632	0.3384	1.9745	0.7632	0.7624	0.0537	0.0773
2.4881	28.0	7000	3.2460	0.7458	0.3618	2.0745	0.7458	0.7479	0.0521	0.0868
2.4881	29.0	7250	3.1299	0.7605	0.3414	1.9781	0.7605	0.7626	0.0600	0.0774
2.469	30.0	7500	3.1695	0.7555	0.3481	2.0246	0.7555	0.7563	0.0534	0.0811
2.469	31.0	7750	3.1766	0.7612	0.3474	1.9997	0.7612	0.7633	0.0541	0.0803
2.4524	32.0	8000	3.0937	0.7638	0.3351	1.9420	0.7638	0.7649	0.0592	0.0754
2.4524	33.0	8250	3.1293	0.7625	0.3409	1.9671	0.7625	0.7633	0.0580	0.0781
2.4382	34.0	8500	3.1129	0.7668	0.3370	2.0003	0.7668	0.7672	0.0623	0.0746
2.4382	35.0	8750	3.0795	0.767	0.3324	1.9772	0.767	0.7677	0.0536	0.0759
2.4259	36.0	9000	3.0927	0.7675	0.3332	1.9557	0.7675	0.7690	0.0588	0.0744
2.4259	37.0	9250	3.0856	0.7702	0.3327	1.9465	0.7702	0.7715	0.0554	0.0756
2.4107	38.0	9500	3.0915	0.7678	0.3319	1.9699	0.7678	0.7681	0.0556	0.0749
2.4107	39.0	9750	3.0885	0.763	0.3338	1.9478	0.763	0.7643	0.0575	0.0750
2.4002	40.0	10000	3.0921	0.771	0.3315	1.9563	0.771	0.7729	0.0557	0.0744
2.4002	41.0	10250	3.0727	0.767	0.3310	1.9530	0.767	0.7682	0.0567	0.0748
2.3905	42.0	10500	3.0793	0.7645	0.3320	1.9484	0.7645	0.7657	0.0598	0.0755
2.3905	43.0	10750	3.0771	0.7672	0.3308	1.9548	0.7672	0.7679	0.0566	0.0737
2.3798	44.0	11000	3.0794	0.7685	0.3309	1.9620	0.7685	0.7693	0.0631	0.0736
2.3798	45.0	11250	3.0736	0.7665	0.3320	1.9408	0.7665	0.7677	0.0589	0.0749
2.3731	46.0	11500	3.0746	0.7682	0.3312	1.9635	0.7682	0.7693	0.0576	0.0743
2.3731	47.0	11750	3.0711	0.768	0.3306	1.9576	0.768	0.7689	0.0572	0.0739
2.3671	48.0	12000	3.0785	0.7682	0.3317	1.9516	0.7682	0.7697	0.0574	0.0744
2.3671	49.0	12250	3.0678	0.7692	0.3298	1.9388	0.7692	0.7700	0.0606	0.0738
2.3628	50.0	12500	3.0688	0.768	0.3296	1.9490	0.768	0.7688	0.0595	0.0737

Framework versions

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