resnet101_rvl-cdip-cnn_rvl_cdip-NK1000_kd

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: 0.6048
• Accuracy: 0.7867
• Brier Loss: 0.3046
• Nll: 2.0167
• F1 Micro: 0.7868
• F1 Macro: 0.7867
• Ece: 0.0468
• Aurc: 0.0597

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	4.1589	0.1305	0.9320	7.8922	0.1305	0.0928	0.0637	0.8337
4.1546	2.0	500	3.6898	0.3515	0.8840	4.7696	0.3515	0.3150	0.2354	0.5486
4.1546	3.0	750	2.3450	0.4863	0.6606	3.2068	0.4863	0.4495	0.0978	0.2927
2.419	4.0	1000	1.5206	0.6125	0.5126	2.7884	0.6125	0.5996	0.0512	0.1677
2.419	5.0	1250	1.2545	0.6593	0.4574	2.6041	0.6593	0.6524	0.0483	0.1337
1.1615	6.0	1500	0.9718	0.704	0.4062	2.4047	0.704	0.7017	0.0506	0.1043
1.1615	7.0	1750	0.8636	0.73	0.3760	2.1975	0.7300	0.7304	0.0522	0.0902
0.7217	8.0	2000	0.7892	0.737	0.3632	2.1583	0.737	0.7377	0.0551	0.0835
0.7217	9.0	2250	0.7438	0.754	0.3470	2.0559	0.754	0.7531	0.0534	0.0766
0.5268	10.0	2500	0.7322	0.758	0.3443	2.1043	0.7580	0.7584	0.0510	0.0742
0.5268	11.0	2750	0.7003	0.7632	0.3335	2.0510	0.7632	0.7639	0.0472	0.0697
0.4197	12.0	3000	0.6921	0.7665	0.3325	2.0569	0.7665	0.7668	0.0568	0.0694
0.4197	13.0	3250	0.7003	0.7618	0.3330	2.0293	0.7618	0.7618	0.0465	0.0721
0.3575	14.0	3500	0.6681	0.7728	0.3244	2.0037	0.7728	0.7739	0.0505	0.0664
0.3575	15.0	3750	0.6862	0.7718	0.3279	2.0294	0.7717	0.7727	0.0442	0.0693
0.3181	16.0	4000	0.6681	0.7738	0.3246	2.0559	0.7738	0.7739	0.0509	0.0671
0.3181	17.0	4250	0.6473	0.7775	0.3177	1.9978	0.7775	0.7784	0.0494	0.0644
0.2874	18.0	4500	0.6448	0.78	0.3172	2.0396	0.78	0.7805	0.0495	0.0651
0.2874	19.0	4750	0.6484	0.779	0.3153	2.0251	0.779	0.7790	0.0519	0.0636
0.2691	20.0	5000	0.6430	0.7768	0.3164	2.0897	0.7768	0.7771	0.0489	0.0635
0.2691	21.0	5250	0.6363	0.78	0.3145	2.0663	0.78	0.7802	0.0476	0.0640
0.2509	22.0	5500	0.6327	0.782	0.3127	2.0358	0.782	0.7820	0.0440	0.0634
0.2509	23.0	5750	0.6287	0.7863	0.3113	2.0157	0.7863	0.7865	0.0463	0.0630
0.2393	24.0	6000	0.6315	0.7778	0.3137	2.0623	0.7778	0.7773	0.0492	0.0633
0.2393	25.0	6250	0.6345	0.7775	0.3149	2.0397	0.7775	0.7773	0.0514	0.0635
0.2291	26.0	6500	0.6233	0.7815	0.3102	1.9988	0.7815	0.7816	0.0444	0.0626
0.2291	27.0	6750	0.6224	0.783	0.3095	2.0085	0.7830	0.7830	0.0502	0.0615
0.2191	28.0	7000	0.6159	0.7835	0.3089	2.0340	0.7835	0.7834	0.0499	0.0614
0.2191	29.0	7250	0.6203	0.7825	0.3096	2.0280	0.7825	0.7825	0.0480	0.0617
0.2139	30.0	7500	0.6233	0.7802	0.3093	2.0660	0.7802	0.7805	0.0518	0.0609
0.2139	31.0	7750	0.6128	0.785	0.3049	2.0148	0.785	0.7851	0.0471	0.0604
0.2068	32.0	8000	0.6124	0.7855	0.3064	2.0336	0.7855	0.7855	0.0433	0.0604
0.2068	33.0	8250	0.6117	0.7835	0.3068	2.0208	0.7835	0.7833	0.0469	0.0604
0.202	34.0	8500	0.6105	0.7857	0.3063	1.9918	0.7857	0.7854	0.0454	0.0611
0.202	35.0	8750	0.6136	0.7877	0.3088	2.0272	0.7877	0.7884	0.0444	0.0607
0.1974	36.0	9000	0.6095	0.786	0.3052	2.0275	0.786	0.7862	0.0423	0.0600
0.1974	37.0	9250	0.6108	0.786	0.3077	2.0035	0.786	0.7860	0.0477	0.0606
0.1945	38.0	9500	0.6107	0.7817	0.3078	2.0042	0.7817	0.7820	0.0482	0.0611
0.1945	39.0	9750	0.6077	0.7875	0.3051	1.9959	0.7875	0.7878	0.0510	0.0599
0.1919	40.0	10000	0.6099	0.7863	0.3072	2.0323	0.7863	0.7866	0.0468	0.0603
0.1919	41.0	10250	0.6046	0.7847	0.3046	2.0113	0.7847	0.7850	0.0442	0.0600
0.1874	42.0	10500	0.6062	0.7865	0.3059	2.0055	0.7865	0.7865	0.0486	0.0598
0.1874	43.0	10750	0.6051	0.787	0.3042	2.0151	0.787	0.7870	0.0451	0.0596
0.1859	44.0	11000	0.6082	0.7855	0.3063	2.0123	0.7855	0.7860	0.0470	0.0600
0.1859	45.0	11250	0.6066	0.7867	0.3047	2.0000	0.7868	0.7865	0.0479	0.0599
0.1856	46.0	11500	0.6049	0.7863	0.3054	2.0058	0.7863	0.7861	0.0475	0.0598
0.1856	47.0	11750	0.6041	0.7887	0.3047	1.9992	0.7887	0.7891	0.0482	0.0595
0.1842	48.0	12000	0.6063	0.7843	0.3055	2.0346	0.7843	0.7843	0.0480	0.0601
0.1842	49.0	12250	0.6058	0.786	0.3051	2.0319	0.786	0.7861	0.0481	0.0598
0.1829	50.0	12500	0.6048	0.7867	0.3046	2.0167	0.7868	0.7867	0.0468	0.0597

Framework versions

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