resnet101_rvl-cdip-cnn_rvl_cdip-NK1000_kd_NKD_t1.0_g1.5

This model is a fine-tuned version of microsoft/resnet-50 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.9013
• Accuracy: 0.7933
• Brier Loss: 0.3080
• Nll: 1.8102
• F1 Micro: 0.7932
• F1 Macro: 0.7937
• Ece: 0.0719
• Aurc: 0.0635

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	6.0054	0.098	0.9327	9.3196	0.0980	0.0481	0.0462	0.8670
6.0141	2.0	500	5.4713	0.2195	0.8933	5.2235	0.2195	0.1452	0.1046	0.7129
6.0141	3.0	750	4.4006	0.4535	0.7034	3.0178	0.4535	0.4351	0.1373	0.3334
4.5079	4.0	1000	3.8431	0.59	0.5686	2.5843	0.59	0.5822	0.1309	0.2072
4.5079	5.0	1250	3.5315	0.6552	0.4864	2.4330	0.6552	0.6537	0.1048	0.1504
3.5028	6.0	1500	3.2850	0.707	0.4163	2.2375	0.707	0.7082	0.0790	0.1111
3.5028	7.0	1750	3.0974	0.7312	0.3721	2.0933	0.7312	0.7328	0.0452	0.0899
3.0599	8.0	2000	3.0385	0.7455	0.3561	2.0148	0.7455	0.7456	0.0432	0.0838
3.0599	9.0	2250	2.9978	0.7565	0.3432	1.9780	0.7565	0.7572	0.0437	0.0777
2.8562	10.0	2500	2.9853	0.7622	0.3397	1.9176	0.7622	0.7619	0.0495	0.0751
2.8562	11.0	2750	2.9803	0.7615	0.3385	1.9327	0.7615	0.7627	0.0547	0.0760
2.7414	12.0	3000	2.9711	0.7658	0.3322	1.9439	0.7658	0.7661	0.0495	0.0740
2.7414	13.0	3250	2.9618	0.771	0.3276	1.8599	0.771	0.7718	0.0548	0.0704
2.6658	14.0	3500	2.9534	0.7762	0.3252	1.8935	0.7762	0.7770	0.0581	0.0699
2.6658	15.0	3750	2.9568	0.776	0.3248	1.8836	0.776	0.7776	0.0588	0.0699
2.6197	16.0	4000	2.9196	0.7812	0.3169	1.8338	0.7812	0.7814	0.0601	0.0655
2.6197	17.0	4250	2.9267	0.7785	0.3202	1.8430	0.7785	0.7783	0.0647	0.0677
2.5794	18.0	4500	2.9189	0.779	0.3155	1.8279	0.779	0.7794	0.0631	0.0661
2.5794	19.0	4750	2.9324	0.7823	0.3177	1.8508	0.7823	0.7823	0.0665	0.0669
2.5553	20.0	5000	2.9192	0.7837	0.3146	1.8312	0.7837	0.7840	0.0641	0.0654
2.5553	21.0	5250	2.9160	0.7817	0.3140	1.8366	0.7817	0.7828	0.0682	0.0658
2.53	22.0	5500	2.9172	0.7837	0.3139	1.8138	0.7837	0.7842	0.0602	0.0652
2.53	23.0	5750	2.9132	0.7875	0.3134	1.8254	0.7875	0.7877	0.0656	0.0646
2.5127	24.0	6000	2.9108	0.7875	0.3132	1.8367	0.7875	0.7869	0.0669	0.0652
2.5127	25.0	6250	2.9272	0.7837	0.3139	1.8551	0.7837	0.7843	0.0632	0.0653
2.4979	26.0	6500	2.9157	0.7867	0.3128	1.8101	0.7868	0.7876	0.0655	0.0647
2.4979	27.0	6750	2.9031	0.785	0.3112	1.8089	0.785	0.7856	0.0688	0.0639
2.4814	28.0	7000	2.9094	0.7875	0.3110	1.8594	0.7875	0.7880	0.0677	0.0646
2.4814	29.0	7250	2.9110	0.7885	0.3116	1.8150	0.7885	0.7891	0.0696	0.0639
2.4741	30.0	7500	2.9039	0.7877	0.3091	1.8471	0.7877	0.7887	0.0694	0.0632
2.4741	31.0	7750	2.9029	0.7907	0.3087	1.7604	0.7907	0.7917	0.0691	0.0633
2.4626	32.0	8000	2.8983	0.7877	0.3094	1.8191	0.7877	0.7884	0.0677	0.0625
2.4626	33.0	8250	2.9024	0.7897	0.3088	1.8025	0.7897	0.7905	0.0720	0.0635
2.4558	34.0	8500	2.9055	0.792	0.3070	1.7869	0.792	0.7920	0.0667	0.0628
2.4558	35.0	8750	2.9055	0.788	0.3104	1.8349	0.788	0.7883	0.0733	0.0645
2.4481	36.0	9000	2.9061	0.7887	0.3078	1.7840	0.7887	0.7898	0.0676	0.0642
2.4481	37.0	9250	2.9086	0.7917	0.3102	1.7942	0.7917	0.7923	0.0716	0.0644
2.4422	38.0	9500	2.9067	0.7897	0.3084	1.7915	0.7897	0.7900	0.0704	0.0637
2.4422	39.0	9750	2.9080	0.7927	0.3092	1.7951	0.7927	0.7930	0.0709	0.0631
2.4386	40.0	10000	2.9064	0.7943	0.3084	1.8079	0.7943	0.7949	0.0734	0.0635
2.4386	41.0	10250	2.8990	0.792	0.3056	1.7918	0.792	0.7924	0.0699	0.0623
2.4312	42.0	10500	2.9057	0.7893	0.3090	1.7892	0.7893	0.7901	0.0735	0.0641
2.4312	43.0	10750	2.8998	0.7923	0.3079	1.7909	0.7923	0.7932	0.0707	0.0630
2.4294	44.0	11000	2.9108	0.7905	0.3090	1.8220	0.7905	0.7916	0.0773	0.0636
2.4294	45.0	11250	2.9030	0.7927	0.3086	1.8126	0.7927	0.7932	0.0710	0.0631
2.4282	46.0	11500	2.9033	0.7915	0.3077	1.8234	0.7915	0.7920	0.0712	0.0631
2.4282	47.0	11750	2.8975	0.7957	0.3063	1.8070	0.7957	0.7968	0.0702	0.0630
2.4246	48.0	12000	2.9049	0.7935	0.3085	1.8090	0.7935	0.7944	0.0722	0.0635
2.4246	49.0	12250	2.9020	0.792	0.3075	1.8233	0.792	0.7927	0.0700	0.0638
2.4227	50.0	12500	2.9013	0.7933	0.3080	1.8102	0.7932	0.7937	0.0719	0.0635

Framework versions

• Transformers 4.33.3
• Pytorch 2.2.0.dev20231002
• Datasets 2.7.1
• Tokenizers 0.13.3