resnet101_rvl-cdip-cnn_rvl_cdip-NK1000_kd_CEKD_t1.0_a0.5_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: 0.5803
• Accuracy: 0.755
• Brier Loss: 0.3388
• Nll: 2.0737
• F1 Micro: 0.755
• F1 Macro: 0.7552
• Ece: 0.0476
• Aurc: 0.0744

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	2.0995	0.2357	0.8533	3.3592	0.2357	0.1996	0.0692	0.5667
2.1408	2.0	500	1.9262	0.309	0.8322	3.1505	0.309	0.2850	0.1062	0.5376
2.1408	3.0	750	1.7386	0.3882	0.7711	2.9990	0.3882	0.3744	0.0840	0.4488
1.5977	4.0	1000	1.7847	0.3885	0.7913	2.9899	0.3885	0.3901	0.1236	0.4675
1.5977	5.0	1250	1.9515	0.3852	0.8032	3.2494	0.3852	0.3563	0.1897	0.4332
1.2396	6.0	1500	1.4812	0.4918	0.6698	2.8748	0.4918	0.4603	0.1135	0.2969
1.2396	7.0	1750	0.9703	0.6342	0.4956	2.5074	0.6342	0.6278	0.0552	0.1583
0.8921	8.0	2000	0.8731	0.6522	0.4638	2.3758	0.6522	0.6529	0.0503	0.1375
0.8921	9.0	2250	0.8524	0.6763	0.4479	2.3526	0.6763	0.6703	0.0533	0.1284
0.6456	10.0	2500	0.8349	0.6753	0.4449	2.3791	0.6753	0.6731	0.0534	0.1266
0.6456	11.0	2750	0.8366	0.6855	0.4410	2.3328	0.6855	0.6840	0.0606	0.1256
0.4631	12.0	3000	0.8733	0.6787	0.4503	2.3609	0.6787	0.6744	0.0814	0.1300
0.4631	13.0	3250	0.8087	0.6877	0.4278	2.3090	0.6877	0.6836	0.0671	0.1149
0.3462	14.0	3500	0.7505	0.7107	0.4037	2.2700	0.7107	0.7112	0.0461	0.1028
0.3462	15.0	3750	0.7638	0.708	0.4031	2.3569	0.708	0.7070	0.0615	0.1001
0.277	16.0	4000	0.8430	0.686	0.4364	2.4615	0.686	0.6876	0.0587	0.1229
0.277	17.0	4250	0.7287	0.7185	0.3946	2.2848	0.7185	0.7172	0.0474	0.0992
0.2452	18.0	4500	0.7114	0.7175	0.3886	2.2763	0.7175	0.7174	0.0551	0.0957
0.2452	19.0	4750	0.6867	0.73	0.3762	2.2263	0.7300	0.7300	0.0483	0.0916
0.2209	20.0	5000	0.7205	0.722	0.3918	2.2900	0.722	0.7235	0.0515	0.0993
0.2209	21.0	5250	0.6912	0.7338	0.3732	2.2467	0.7338	0.7333	0.0505	0.0892
0.2048	22.0	5500	0.7503	0.7073	0.4025	2.3059	0.7073	0.7067	0.0517	0.1025
0.2048	23.0	5750	0.6887	0.724	0.3822	2.2328	0.724	0.7249	0.0470	0.0930
0.1941	24.0	6000	0.6539	0.7345	0.3664	2.1801	0.7345	0.7348	0.0479	0.0853
0.1941	25.0	6250	0.6466	0.7362	0.3612	2.1945	0.7362	0.7343	0.0501	0.0827
0.1836	26.0	6500	0.6709	0.731	0.3738	2.2298	0.731	0.7351	0.0484	0.0900
0.1836	27.0	6750	0.6462	0.7365	0.3660	2.1816	0.7365	0.7392	0.0498	0.0861
0.1753	28.0	7000	0.6312	0.746	0.3580	2.1495	0.746	0.7469	0.0465	0.0816
0.1753	29.0	7250	0.6152	0.7518	0.3508	2.1226	0.7518	0.7514	0.0390	0.0788
0.1682	30.0	7500	0.6555	0.737	0.3671	2.1765	0.737	0.7373	0.0412	0.0856
0.1682	31.0	7750	0.6287	0.741	0.3565	2.1122	0.7410	0.7419	0.0438	0.0807
0.1622	32.0	8000	0.6226	0.7442	0.3571	2.1540	0.7442	0.7426	0.0445	0.0821
0.1622	33.0	8250	0.6171	0.748	0.3522	2.1308	0.748	0.7482	0.0412	0.0793
0.1565	34.0	8500	0.6127	0.7445	0.3523	2.1276	0.7445	0.7434	0.0402	0.0800
0.1565	35.0	8750	0.6059	0.7482	0.3476	2.1240	0.7482	0.7481	0.0471	0.0770
0.1513	36.0	9000	0.6063	0.7462	0.3493	2.0960	0.7462	0.7472	0.0448	0.0790
0.1513	37.0	9250	0.6078	0.746	0.3513	2.1283	0.746	0.7461	0.0433	0.0792
0.1467	38.0	9500	0.5969	0.753	0.3448	2.1077	0.753	0.7538	0.0364	0.0776
0.1467	39.0	9750	0.6008	0.7475	0.3474	2.1154	0.7475	0.7485	0.0410	0.0794
0.1424	40.0	10000	0.5998	0.7518	0.3464	2.0894	0.7518	0.7526	0.0440	0.0783
0.1424	41.0	10250	0.5904	0.7508	0.3422	2.0606	0.7508	0.7505	0.0455	0.0758
0.1384	42.0	10500	0.5890	0.7552	0.3410	2.0815	0.7552	0.7548	0.0362	0.0744
0.1384	43.0	10750	0.5847	0.7562	0.3404	2.0912	0.7562	0.7568	0.0417	0.0747
0.1348	44.0	11000	0.5856	0.7528	0.3409	2.0675	0.7528	0.7525	0.0431	0.0753
0.1348	45.0	11250	0.5859	0.7552	0.3409	2.0913	0.7552	0.7551	0.0413	0.0750
0.1319	46.0	11500	0.5810	0.7528	0.3391	2.0654	0.7528	0.7526	0.0396	0.0747
0.1319	47.0	11750	0.5803	0.7535	0.3387	2.0705	0.7535	0.7534	0.0444	0.0746
0.13	48.0	12000	0.5819	0.7532	0.3393	2.0563	0.7532	0.7540	0.0390	0.0749
0.13	49.0	12250	0.5800	0.756	0.3383	2.0649	0.756	0.7562	0.0370	0.0740
0.1286	50.0	12500	0.5803	0.755	0.3388	2.0737	0.755	0.7552	0.0476	0.0744

Framework versions

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