INTERNAL_BEST-safety-utcustom-train-SF-RGB-b5

This model is a fine-tuned version of nvidia/mit-b5 on the sam1120/safety-utcustom-TRAIN dataset. It achieves the following results on the evaluation set:

• Loss: 0.0472
• Mean Iou: 0.8728
• Mean Accuracy: 0.9195
• Overall Accuracy: 0.9919
• Accuracy Unlabeled: nan
• Accuracy Safe: 0.8426
• Accuracy Unsafe: 0.9964
• Iou Unlabeled: nan
• Iou Safe: 0.7540
• Iou Unsafe: 0.9917

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Safe	Accuracy Unsafe	Iou Unlabeled	Iou Safe	Iou Unsafe
1.2199	2.0	20	1.1474	0.0765	0.4399	0.2168	nan	0.6770	0.2028	0.0	0.0279	0.2014
1.1542	4.0	40	1.0616	0.1558	0.6082	0.4365	nan	0.7908	0.4257	0.0	0.0436	0.4237
1.0324	6.0	60	0.9140	0.2569	0.6886	0.7091	nan	0.6667	0.7104	0.0	0.0672	0.7036
0.8058	8.0	80	0.7629	0.2970	0.6780	0.8115	nan	0.5360	0.8199	0.0	0.0823	0.8086
0.681	10.0	100	0.5545	0.3510	0.6964	0.9135	nan	0.4657	0.9271	0.0	0.1407	0.9123
0.5248	12.0	120	0.4153	0.3738	0.6747	0.9462	nan	0.3861	0.9633	0.0	0.1757	0.9456
0.3372	14.0	140	0.3050	0.3896	0.6548	0.9628	nan	0.3276	0.9821	0.0	0.2064	0.9624
0.2818	16.0	160	0.2346	0.4313	0.7331	0.9703	nan	0.4810	0.9852	0.0	0.3239	0.9699
0.2081	18.0	180	0.1802	0.4762	0.7973	0.9782	nan	0.6050	0.9896	0.0	0.4509	0.9778
0.141	20.0	200	0.1362	0.4968	0.7945	0.9830	nan	0.5943	0.9948	0.0	0.5078	0.9827
0.0963	22.0	220	0.1032	0.7409	0.7652	0.9841	nan	0.5326	0.9979	nan	0.4978	0.9839
0.0842	24.0	240	0.0770	0.7840	0.8200	0.9863	nan	0.6432	0.9968	nan	0.5818	0.9861
0.0702	26.0	260	0.0669	0.7836	0.8193	0.9863	nan	0.6417	0.9968	nan	0.5812	0.9861
0.0706	28.0	280	0.0671	0.8065	0.8593	0.9872	nan	0.7234	0.9953	nan	0.6261	0.9870
0.0747	30.0	300	0.0551	0.7808	0.7980	0.9870	nan	0.5971	0.9988	nan	0.5748	0.9867
0.057	32.0	320	0.0492	0.8267	0.8736	0.9888	nan	0.7511	0.9961	nan	0.6648	0.9886
0.0435	34.0	340	0.0507	0.7956	0.8134	0.9878	nan	0.6280	0.9988	nan	0.6035	0.9876
0.0326	36.0	360	0.0418	0.8422	0.8895	0.9898	nan	0.7830	0.9961	nan	0.6947	0.9896
0.0262	38.0	380	0.0420	0.8280	0.8550	0.9895	nan	0.7120	0.9979	nan	0.6667	0.9893
0.0268	40.0	400	0.0392	0.8407	0.8822	0.9899	nan	0.7676	0.9967	nan	0.6918	0.9897
0.0395	42.0	420	0.0466	0.5436	0.8370	0.9889	nan	0.6755	0.9984	0.0	0.6422	0.9887
0.0279	44.0	440	0.0439	0.8321	0.8946	0.9887	nan	0.7945	0.9947	nan	0.6758	0.9885
0.0468	46.0	460	0.0360	0.8480	0.8894	0.9904	nan	0.7822	0.9967	nan	0.7059	0.9901
0.0233	48.0	480	0.0376	0.8507	0.8962	0.9905	nan	0.7960	0.9964	nan	0.7113	0.9902
0.0288	50.0	500	0.0386	0.8404	0.8845	0.9898	nan	0.7725	0.9964	nan	0.6913	0.9896
0.0266	52.0	520	0.0361	0.8455	0.8768	0.9905	nan	0.7560	0.9976	nan	0.7008	0.9902
0.0241	54.0	540	0.0367	0.8504	0.9039	0.9902	nan	0.8122	0.9957	nan	0.7109	0.9900
0.0239	56.0	560	0.0414	0.8401	0.8809	0.9899	nan	0.7650	0.9967	nan	0.6906	0.9896
0.0221	58.0	580	0.0375	0.8536	0.8971	0.9907	nan	0.7977	0.9966	nan	0.7167	0.9905
0.0324	60.0	600	0.0390	0.8566	0.9017	0.9908	nan	0.8069	0.9964	nan	0.7226	0.9906
0.0264	62.0	620	0.0405	0.8506	0.9088	0.9901	nan	0.8224	0.9952	nan	0.7114	0.9899
0.0146	64.0	640	0.0356	0.8627	0.9158	0.9911	nan	0.8358	0.9958	nan	0.7346	0.9909
0.0252	66.0	660	0.0310	0.8667	0.9010	0.9917	nan	0.8046	0.9974	nan	0.7418	0.9915
0.0155	68.0	680	0.0359	0.8567	0.9056	0.9908	nan	0.8152	0.9961	nan	0.7229	0.9905
0.0169	70.0	700	0.0490	0.8476	0.9182	0.9896	nan	0.8422	0.9941	nan	0.7058	0.9894
0.0142	72.0	720	0.0357	0.8442	0.8771	0.9903	nan	0.7568	0.9974	nan	0.6982	0.9901
0.0244	74.0	740	0.0400	0.8523	0.9070	0.9903	nan	0.8183	0.9956	nan	0.7146	0.9901
0.016	76.0	760	0.0302	0.8644	0.9037	0.9915	nan	0.8105	0.9970	nan	0.7376	0.9913
0.0137	78.0	780	0.0325	0.8664	0.9118	0.9915	nan	0.8271	0.9965	nan	0.7415	0.9913
0.0115	80.0	800	0.0347	0.8678	0.9162	0.9915	nan	0.8362	0.9962	nan	0.7443	0.9913
0.0117	82.0	820	0.0320	0.8697	0.9084	0.9918	nan	0.8197	0.9971	nan	0.7478	0.9916
0.0108	84.0	840	0.0348	0.8691	0.9192	0.9916	nan	0.8423	0.9961	nan	0.7468	0.9913
0.0101	86.0	860	0.0342	0.8683	0.9081	0.9917	nan	0.8192	0.9970	nan	0.7452	0.9915
0.0085	88.0	880	0.0441	0.8639	0.9214	0.9911	nan	0.8474	0.9954	nan	0.7370	0.9908
0.009	90.0	900	0.0428	0.8619	0.9086	0.9912	nan	0.8209	0.9963	nan	0.7330	0.9909
0.009	92.0	920	0.0444	0.8620	0.9089	0.9912	nan	0.8215	0.9963	nan	0.7331	0.9909
0.0089	94.0	940	0.0410	0.8645	0.9147	0.9913	nan	0.8334	0.9961	nan	0.7380	0.9910
0.0091	96.0	960	0.0418	0.8663	0.9155	0.9914	nan	0.8349	0.9962	nan	0.7413	0.9912
0.0079	98.0	980	0.0398	0.8629	0.9085	0.9913	nan	0.8205	0.9965	nan	0.7348	0.9910
0.0084	100.0	1000	0.0497	0.8553	0.9109	0.9905	nan	0.8262	0.9955	nan	0.7204	0.9903
0.0088	102.0	1020	0.0399	0.8558	0.9058	0.9907	nan	0.8156	0.9960	nan	0.7212	0.9905
0.0089	104.0	1040	0.0388	0.8678	0.9225	0.9914	nan	0.8494	0.9957	nan	0.7444	0.9912
0.008	106.0	1060	0.0449	0.8622	0.9225	0.9909	nan	0.8498	0.9952	nan	0.7337	0.9907
0.0084	108.0	1080	0.0429	0.8687	0.9233	0.9914	nan	0.8510	0.9957	nan	0.7462	0.9912
0.0084	110.0	1100	0.0405	0.8687	0.9169	0.9916	nan	0.8375	0.9963	nan	0.7460	0.9914
0.007	112.0	1120	0.0544	0.8620	0.9180	0.9910	nan	0.8404	0.9956	nan	0.7333	0.9907
0.0079	114.0	1140	0.0501	0.8602	0.9176	0.9908	nan	0.8399	0.9954	nan	0.7299	0.9906
0.0084	116.0	1160	0.0508	0.8605	0.9212	0.9908	nan	0.8473	0.9951	nan	0.7304	0.9905
0.0113	118.0	1180	0.0511	0.8601	0.9228	0.9907	nan	0.8507	0.9950	nan	0.7298	0.9905
0.0076	120.0	1200	0.0556	0.8602	0.9264	0.9906	nan	0.8582	0.9947	nan	0.7299	0.9904
0.0081	122.0	1220	0.0471	0.8665	0.9256	0.9912	nan	0.8559	0.9953	nan	0.7420	0.9910
0.0054	124.0	1240	0.0504	0.8652	0.9174	0.9913	nan	0.8389	0.9959	nan	0.7394	0.9910
0.0054	126.0	1260	0.0502	0.8666	0.9209	0.9913	nan	0.8461	0.9957	nan	0.7420	0.9911
0.0092	128.0	1280	0.0540	0.8642	0.9223	0.9911	nan	0.8492	0.9954	nan	0.7376	0.9908
0.007	130.0	1300	0.0533	0.8637	0.9207	0.9911	nan	0.8459	0.9955	nan	0.7366	0.9908
0.0063	132.0	1320	0.0526	0.8644	0.9259	0.9910	nan	0.8567	0.9951	nan	0.7380	0.9908
0.0101	134.0	1340	0.0462	0.8653	0.9067	0.9915	nan	0.8166	0.9968	nan	0.7393	0.9913
0.0183	136.0	1360	0.0516	0.5675	0.9024	0.9904	nan	0.8089	0.9959	0.0	0.7123	0.9901
0.0102	138.0	1380	0.0388	0.8366	0.8716	0.9898	nan	0.7460	0.9972	nan	0.6837	0.9896
0.0277	140.0	1400	0.0649	0.8159	0.9590	0.9850	nan	0.9313	0.9866	nan	0.6472	0.9846
0.0169	142.0	1420	0.0340	0.8444	0.9148	0.9894	nan	0.8355	0.9941	nan	0.6996	0.9891
0.0359	144.0	1440	0.0314	0.8667	0.8987	0.9918	nan	0.7997	0.9976	nan	0.7419	0.9916
0.0117	146.0	1460	0.0307	0.8517	0.8869	0.9908	nan	0.7765	0.9973	nan	0.7129	0.9905
0.0097	148.0	1480	0.0323	0.8757	0.9070	0.9924	nan	0.8162	0.9977	nan	0.7593	0.9922
0.0063	150.0	1500	0.0302	0.8808	0.9155	0.9926	nan	0.8335	0.9975	nan	0.7692	0.9924
0.0085	152.0	1520	0.0352	0.8697	0.9113	0.9918	nan	0.8258	0.9968	nan	0.7478	0.9916
0.0078	154.0	1540	0.0428	0.8649	0.9190	0.9912	nan	0.8423	0.9957	nan	0.7388	0.9910
0.0056	156.0	1560	0.0340	0.8709	0.9170	0.9918	nan	0.8376	0.9965	nan	0.7502	0.9916
0.0063	158.0	1580	0.0359	0.8661	0.9201	0.9913	nan	0.8445	0.9958	nan	0.7412	0.9911
0.0083	160.0	1600	0.0375	0.8684	0.9186	0.9915	nan	0.8410	0.9961	nan	0.7456	0.9913
0.0065	162.0	1620	0.0370	0.8699	0.9210	0.9916	nan	0.8459	0.9960	nan	0.7484	0.9914
0.0063	164.0	1640	0.0388	0.8699	0.9228	0.9916	nan	0.8498	0.9959	nan	0.7484	0.9913
0.0056	166.0	1660	0.0386	0.8702	0.9238	0.9916	nan	0.8517	0.9958	nan	0.7491	0.9914
0.0049	168.0	1680	0.0394	0.8703	0.9199	0.9917	nan	0.8436	0.9962	nan	0.7491	0.9914
0.0054	170.0	1700	0.0400	0.8704	0.9195	0.9917	nan	0.8428	0.9962	nan	0.7494	0.9915
0.0046	172.0	1720	0.0398	0.8728	0.9187	0.9919	nan	0.8410	0.9965	nan	0.7539	0.9917
0.0058	174.0	1740	0.0402	0.8711	0.9166	0.9918	nan	0.8367	0.9965	nan	0.7507	0.9916
0.005	176.0	1760	0.0400	0.8720	0.9196	0.9918	nan	0.8428	0.9963	nan	0.7525	0.9916
0.0061	178.0	1780	0.0417	0.8714	0.9226	0.9917	nan	0.8492	0.9960	nan	0.7513	0.9915
0.0061	180.0	1800	0.0407	0.8731	0.9249	0.9918	nan	0.8538	0.9960	nan	0.7545	0.9916
0.0065	182.0	1820	0.0420	0.8712	0.9235	0.9917	nan	0.8511	0.9959	nan	0.7509	0.9914
0.0045	184.0	1840	0.0421	0.8718	0.9250	0.9917	nan	0.8541	0.9959	nan	0.7522	0.9915
0.0056	186.0	1860	0.0435	0.8703	0.9157	0.9917	nan	0.8349	0.9965	nan	0.7490	0.9915
0.0059	188.0	1880	0.0436	0.8707	0.9191	0.9917	nan	0.8419	0.9963	nan	0.7499	0.9915
0.0042	190.0	1900	0.0436	0.8707	0.9210	0.9917	nan	0.8458	0.9961	nan	0.7499	0.9915
0.006	192.0	1920	0.0426	0.8697	0.9193	0.9916	nan	0.8425	0.9962	nan	0.7480	0.9914
0.0053	194.0	1940	0.0447	0.8697	0.9199	0.9916	nan	0.8437	0.9961	nan	0.7480	0.9914
0.0044	196.0	1960	0.0441	0.8710	0.9238	0.9916	nan	0.8516	0.9959	nan	0.7505	0.9914
0.0049	198.0	1980	0.0453	0.8693	0.9219	0.9915	nan	0.8479	0.9959	nan	0.7473	0.9913
0.0059	200.0	2000	0.0444	0.8726	0.9233	0.9918	nan	0.8506	0.9961	nan	0.7537	0.9916
0.005	202.0	2020	0.0447	0.8717	0.9256	0.9917	nan	0.8555	0.9958	nan	0.7519	0.9914
0.005	204.0	2040	0.0451	0.8711	0.9227	0.9917	nan	0.8494	0.9960	nan	0.7507	0.9915
0.0043	206.0	2060	0.0458	0.8707	0.9220	0.9916	nan	0.8480	0.9960	nan	0.7499	0.9914
0.0043	208.0	2080	0.0462	0.8696	0.9221	0.9916	nan	0.8482	0.9959	nan	0.7479	0.9913
0.0062	210.0	2100	0.0451	0.8715	0.9211	0.9917	nan	0.8461	0.9962	nan	0.7515	0.9915
0.0056	212.0	2120	0.0470	0.8706	0.9213	0.9917	nan	0.8466	0.9961	nan	0.7498	0.9914
0.0049	214.0	2140	0.0480	0.8679	0.9229	0.9914	nan	0.8500	0.9957	nan	0.7447	0.9912
0.0038	216.0	2160	0.0474	0.8700	0.9194	0.9916	nan	0.8427	0.9962	nan	0.7486	0.9914
0.0043	218.0	2180	0.0472	0.8693	0.9231	0.9915	nan	0.8503	0.9958	nan	0.7474	0.9913
0.005	220.0	2200	0.0471	0.8695	0.9156	0.9917	nan	0.8348	0.9964	nan	0.7475	0.9915
0.0041	222.0	2220	0.0472	0.8719	0.9187	0.9918	nan	0.8411	0.9964	nan	0.7522	0.9916
0.0041	224.0	2240	0.0471	0.8720	0.9219	0.9918	nan	0.8477	0.9962	nan	0.7525	0.9916
0.005	226.0	2260	0.0479	0.8720	0.9191	0.9918	nan	0.8418	0.9964	nan	0.7524	0.9916
0.0041	228.0	2280	0.0489	0.8706	0.9183	0.9917	nan	0.8403	0.9963	nan	0.7498	0.9915
0.005	230.0	2300	0.0472	0.8728	0.9195	0.9919	nan	0.8426	0.9964	nan	0.7540	0.9917

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3