drone_DinoVdeau-large-2024_09_18-batch-size64_epochs100_freeze

This model is a fine-tuned version of facebook/dinov2-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2981
• F1 Micro: 0.8388
• F1 Macro: 0.6139
• Roc Auc: 0.8667
• Accuracy: 0.2677
• Learning Rate: 1e-05

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: 0.001
• 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
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Roc Auc	Accuracy	Rate
No log	1.0	181	0.3341	0.8203	0.5745	0.8508	0.2409	0.001
No log	2.0	362	0.3186	0.8327	0.5964	0.8618	0.2401	0.001
0.41	3.0	543	0.3190	0.8323	0.5896	0.8611	0.2516	0.001
0.41	4.0	724	0.3149	0.8322	0.5800	0.8608	0.2573	0.001
0.41	5.0	905	0.3153	0.8314	0.5992	0.8601	0.2526	0.001
0.3412	6.0	1086	0.3147	0.8368	0.6059	0.8655	0.2505	0.001
0.3412	7.0	1267	0.3130	0.8297	0.5672	0.8583	0.2643	0.001
0.3412	8.0	1448	0.3127	0.8305	0.5794	0.8590	0.2643	0.001
0.3338	9.0	1629	0.3131	0.8327	0.5719	0.8608	0.2700	0.001
0.3338	10.0	1810	0.3097	0.8355	0.5897	0.8638	0.2518	0.001
0.3338	11.0	1991	0.3123	0.8332	0.5735	0.8612	0.2742	0.001
0.3303	12.0	2172	0.3085	0.8331	0.5806	0.8612	0.2706	0.001
0.3303	13.0	2353	0.3079	0.8348	0.5952	0.8628	0.2646	0.001
0.3278	14.0	2534	0.3165	0.8340	0.5969	0.8626	0.2534	0.001
0.3278	15.0	2715	0.3074	0.8351	0.5790	0.8631	0.2760	0.001
0.3278	16.0	2896	0.3095	0.8356	0.5889	0.8637	0.2635	0.001
0.3273	17.0	3077	0.3103	0.8395	0.6138	0.8680	0.2474	0.001
0.3273	18.0	3258	0.3063	0.8334	0.5715	0.8611	0.2763	0.001
0.3273	19.0	3439	0.3110	0.8337	0.5920	0.8617	0.2656	0.001
0.324	20.0	3620	0.3072	0.8375	0.5984	0.8655	0.2596	0.001
0.324	21.0	3801	0.3074	0.8389	0.6090	0.8672	0.2560	0.001
0.324	22.0	3982	0.3070	0.8355	0.5808	0.8634	0.2656	0.001
0.3263	23.0	4163	0.3077	0.8389	0.6160	0.8669	0.2627	0.001
0.3263	24.0	4344	0.3061	0.8363	0.5881	0.8640	0.2656	0.001
0.3244	25.0	4525	0.3043	0.8402	0.6102	0.8679	0.2664	0.001
0.3244	26.0	4706	0.3110	0.8327	0.5806	0.8610	0.2659	0.001
0.3244	27.0	4887	0.3052	0.8380	0.5850	0.8656	0.2713	0.001
0.3257	28.0	5068	0.3030	0.8397	0.5974	0.8674	0.2664	0.001
0.3257	29.0	5249	0.3067	0.8362	0.5902	0.8642	0.2666	0.001
0.3257	30.0	5430	0.3061	0.8363	0.5924	0.8644	0.2635	0.001
0.3243	31.0	5611	0.3028	0.8373	0.5867	0.8649	0.2708	0.001
0.3243	32.0	5792	0.3060	0.8388	0.6094	0.8667	0.2568	0.001
0.3243	33.0	5973	0.3069	0.8342	0.5866	0.8625	0.2651	0.001
0.3257	34.0	6154	0.3069	0.8363	0.5901	0.8641	0.2664	0.001
0.3257	35.0	6335	0.3041	0.8380	0.6009	0.8657	0.2627	0.001
0.324	36.0	6516	0.3045	0.8363	0.5947	0.8640	0.2661	0.001
0.324	37.0	6697	0.3037	0.8396	0.5995	0.8672	0.2760	0.001
0.324	38.0	6878	0.3015	0.8388	0.5860	0.8662	0.2737	0.0001
0.3203	39.0	7059	0.3005	0.8381	0.5995	0.8656	0.2737	0.0001
0.3203	40.0	7240	0.3010	0.8417	0.6126	0.8692	0.2695	0.0001
0.3203	41.0	7421	0.2990	0.8403	0.6073	0.8677	0.2742	0.0001
0.3165	42.0	7602	0.2996	0.8409	0.5992	0.8681	0.2713	0.0001
0.3165	43.0	7783	0.2986	0.8414	0.6092	0.8688	0.2695	0.0001
0.3165	44.0	7964	0.2982	0.8396	0.5954	0.8668	0.2750	0.0001
0.3138	45.0	8145	0.2977	0.8401	0.6028	0.8674	0.2758	0.0001
0.3138	46.0	8326	0.2982	0.8406	0.5966	0.8677	0.2755	0.0001
0.3125	47.0	8507	0.2997	0.8378	0.5893	0.8650	0.2768	0.0001
0.3125	48.0	8688	0.2978	0.8420	0.6135	0.8694	0.2747	0.0001
0.3125	49.0	8869	0.2981	0.8399	0.6017	0.8671	0.2747	0.0001
0.312	50.0	9050	0.2977	0.8410	0.6113	0.8684	0.2703	0.0001
0.312	51.0	9231	0.3001	0.8419	0.6110	0.8696	0.2716	0.0001
0.312	52.0	9412	0.2977	0.8380	0.6011	0.8653	0.2789	0.0001
0.3115	53.0	9593	0.2966	0.8425	0.6151	0.8699	0.2729	0.0001
0.3115	54.0	9774	0.2977	0.8399	0.5974	0.8669	0.2810	0.0001
0.3115	55.0	9955	0.2965	0.8407	0.6119	0.8680	0.2724	0.0001
0.3105	56.0	10136	0.2966	0.8408	0.6058	0.8679	0.2786	0.0001
0.3105	57.0	10317	0.2978	0.8399	0.6068	0.8672	0.2747	0.0001
0.3105	58.0	10498	0.2965	0.8427	0.6146	0.8699	0.2721	0.0001
0.3105	59.0	10679	0.2961	0.8406	0.6006	0.8676	0.2794	0.0001
0.3105	60.0	10860	0.2961	0.8429	0.6113	0.8700	0.2797	0.0001
0.308	61.0	11041	0.2963	0.8415	0.5999	0.8684	0.2797	0.0001
0.308	62.0	11222	0.2961	0.8405	0.6017	0.8676	0.2789	0.0001
0.308	63.0	11403	0.2955	0.8421	0.6108	0.8693	0.2737	0.0001
0.3083	64.0	11584	0.2952	0.8407	0.6127	0.8679	0.2791	0.0001
0.3083	65.0	11765	0.2984	0.8391	0.6022	0.8664	0.2781	0.0001
0.3083	66.0	11946	0.2957	0.8415	0.6104	0.8687	0.2739	0.0001
0.3051	67.0	12127	0.2962	0.8416	0.6142	0.8690	0.2760	0.0001
0.3051	68.0	12308	0.2967	0.8413	0.6084	0.8686	0.2773	0.0001
0.3051	69.0	12489	0.2960	0.8406	0.6161	0.8679	0.2729	0.0001
0.3066	70.0	12670	0.2972	0.8434	0.6248	0.8708	0.2682	0.0001
0.3066	71.0	12851	0.2965	0.8396	0.6081	0.8668	0.2802	1e-05
0.3061	72.0	13032	0.2961	0.8422	0.6034	0.8691	0.2807	1e-05
0.3061	73.0	13213	0.2954	0.8409	0.6080	0.8679	0.2810	1e-05
0.3061	74.0	13394	0.2954	0.8411	0.6093	0.8682	0.2810	1e-05

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1