INTERNAL_BEST-safety-utcustom-train-SF-RGBD-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: 0.0506
• Mean Iou: 0.8519
• Mean Accuracy: 0.9125
• Overall Accuracy: 0.9902
• Accuracy Unlabeled: nan
• Accuracy Safe: 0.8300
• Accuracy Unsafe: 0.9950
• Iou Unlabeled: nan
• Iou Safe: 0.7138
• Iou Unsafe: 0.9899

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
1.4012	2.0	20	1.1653	0.0379	0.0923	0.0926	nan	0.0920	0.0926	0.0	0.0214	0.0923
1.2461	4.0	40	0.9899	0.2255	0.3670	0.6307	nan	0.0868	0.6473	0.0	0.0379	0.6386
1.0596	6.0	60	0.7738	0.2703	0.4188	0.7941	nan	0.0199	0.8177	0.0	0.0143	0.7967
0.8267	8.0	80	0.6767	0.2902	0.4512	0.8507	nan	0.0265	0.8758	0.0	0.0183	0.8522
0.7282	10.0	100	0.5637	0.3086	0.4776	0.9098	nan	0.0183	0.9370	0.0	0.0149	0.9110
0.5124	12.0	120	0.4667	0.3254	0.5053	0.9512	nan	0.0314	0.9792	0.0	0.0247	0.9516
0.3126	14.0	140	0.3585	0.3325	0.5147	0.9662	nan	0.0349	0.9945	0.0	0.0313	0.9663
0.2862	16.0	160	0.2890	0.3346	0.5168	0.9703	nan	0.0349	0.9988	0.0	0.0336	0.9703
0.2374	18.0	180	0.2102	0.3647	0.5637	0.9725	nan	0.1291	0.9982	0.0	0.1218	0.9724
0.1583	20.0	200	0.1730	0.6293	0.6574	0.9761	nan	0.3186	0.9961	nan	0.2827	0.9759
0.1082	22.0	220	0.1317	0.6306	0.6566	0.9765	nan	0.3166	0.9966	nan	0.2849	0.9763
0.1025	24.0	240	0.1116	0.6494	0.6766	0.9777	nan	0.3565	0.9967	nan	0.3212	0.9775
0.1158	26.0	260	0.0965	0.7200	0.7978	0.9791	nan	0.6051	0.9905	nan	0.4612	0.9787
0.0882	28.0	280	0.0857	0.7356	0.7857	0.9822	nan	0.5769	0.9946	nan	0.4893	0.9819
0.07	30.0	300	0.0829	0.6717	0.6934	0.9799	nan	0.3890	0.9979	nan	0.3637	0.9797
0.0911	32.0	320	0.0677	0.7680	0.8244	0.9843	nan	0.6545	0.9944	nan	0.5521	0.9840
0.0807	34.0	340	0.0696	0.7779	0.8716	0.9834	nan	0.7528	0.9904	nan	0.5727	0.9830
0.0531	36.0	360	0.0611	0.7761	0.8781	0.9829	nan	0.7668	0.9895	nan	0.5698	0.9825
0.0407	38.0	380	0.0567	0.7828	0.8396	0.9854	nan	0.6846	0.9945	nan	0.5805	0.9851
0.0449	40.0	400	0.0639	0.7725	0.8200	0.9851	nan	0.6446	0.9954	nan	0.5602	0.9848
0.0932	42.0	420	0.0503	0.7726	0.7983	0.9861	nan	0.5987	0.9979	nan	0.5593	0.9858
0.0362	44.0	440	0.0634	0.7553	0.8670	0.9805	nan	0.7464	0.9876	nan	0.5306	0.9801
0.0324	46.0	460	0.0501	0.8024	0.8615	0.9867	nan	0.7284	0.9946	nan	0.6184	0.9864
0.036	48.0	480	0.0454	0.8010	0.8454	0.9872	nan	0.6947	0.9961	nan	0.6151	0.9869
0.0356	50.0	500	0.0495	0.8061	0.8760	0.9866	nan	0.7585	0.9936	nan	0.6260	0.9863
0.0333	52.0	520	0.0483	0.7743	0.8128	0.9856	nan	0.6292	0.9964	nan	0.5632	0.9853
0.0277	54.0	540	0.0445	0.7714	0.7932	0.9862	nan	0.5880	0.9983	nan	0.5569	0.9859
0.0298	56.0	560	0.0460	0.8034	0.8518	0.9872	nan	0.7078	0.9957	nan	0.6198	0.9869
0.0256	58.0	580	0.0416	0.8181	0.8548	0.9886	nan	0.7126	0.9970	nan	0.6479	0.9883
0.0336	60.0	600	0.0442	0.7957	0.8168	0.9877	nan	0.6351	0.9984	nan	0.6039	0.9875
0.0283	62.0	620	0.0425	0.8141	0.8812	0.9873	nan	0.7684	0.9940	nan	0.6413	0.9870
0.0198	64.0	640	0.0455	0.8059	0.8401	0.9879	nan	0.6830	0.9971	nan	0.6242	0.9876
0.0181	66.0	660	0.0444	0.8144	0.8733	0.9876	nan	0.7519	0.9948	nan	0.6415	0.9873
0.0188	68.0	680	0.0456	0.8179	0.8696	0.9881	nan	0.7436	0.9955	nan	0.6479	0.9878
0.0165	70.0	700	0.0431	0.8208	0.8985	0.9875	nan	0.8040	0.9930	nan	0.6544	0.9872
0.0184	72.0	720	0.0421	0.8165	0.8785	0.9876	nan	0.7625	0.9945	nan	0.6457	0.9874
0.0336	74.0	740	0.0441	0.8081	0.8792	0.9867	nan	0.7650	0.9935	nan	0.6298	0.9864
0.0165	76.0	760	0.0374	0.8200	0.8555	0.9887	nan	0.7139	0.9971	nan	0.6515	0.9885
0.0127	78.0	780	0.0402	0.8222	0.8780	0.9882	nan	0.7608	0.9952	nan	0.6563	0.9880
0.0152	80.0	800	0.0430	0.8230	0.8687	0.9886	nan	0.7413	0.9961	nan	0.6576	0.9883
0.0143	82.0	820	0.0410	0.8087	0.8422	0.9881	nan	0.6873	0.9972	nan	0.6297	0.9878
0.0134	84.0	840	0.0335	0.8429	0.8893	0.9899	nan	0.7823	0.9962	nan	0.6962	0.9897
0.0122	86.0	860	0.0396	0.8312	0.8749	0.9892	nan	0.7534	0.9964	nan	0.6734	0.9890
0.0126	88.0	880	0.0405	0.8341	0.8805	0.9893	nan	0.7649	0.9962	nan	0.6791	0.9891
0.0121	90.0	900	0.0400	0.8390	0.8810	0.9898	nan	0.7654	0.9966	nan	0.6884	0.9895
0.0104	92.0	920	0.0372	0.8453	0.8990	0.9899	nan	0.8024	0.9956	nan	0.7010	0.9896
0.0128	94.0	940	0.0394	0.8411	0.8893	0.9897	nan	0.7825	0.9961	nan	0.6927	0.9895
0.0124	96.0	960	0.0409	0.8395	0.8948	0.9895	nan	0.7943	0.9954	nan	0.6899	0.9892
0.0095	98.0	980	0.0413	0.8258	0.8903	0.9882	nan	0.7863	0.9944	nan	0.6637	0.9880
0.0147	100.0	1000	0.0468	0.8181	0.9044	0.9870	nan	0.8167	0.9922	nan	0.6496	0.9867
0.0125	102.0	1020	0.0379	0.8213	0.8961	0.9876	nan	0.7989	0.9933	nan	0.6553	0.9873
0.0142	104.0	1040	0.0328	0.8449	0.9154	0.9894	nan	0.8366	0.9941	nan	0.7006	0.9892
0.0101	106.0	1060	0.0428	0.8407	0.9144	0.9891	nan	0.8351	0.9937	nan	0.6927	0.9888
0.0097	108.0	1080	0.0397	0.8296	0.8847	0.9888	nan	0.7740	0.9953	nan	0.6707	0.9885
0.01	110.0	1100	0.0384	0.8457	0.8935	0.9901	nan	0.7910	0.9961	nan	0.7016	0.9898
0.0084	112.0	1120	0.0385	0.8421	0.8874	0.9899	nan	0.7784	0.9963	nan	0.6945	0.9896
0.0086	114.0	1140	0.0413	0.8488	0.8882	0.9905	nan	0.7795	0.9969	nan	0.7074	0.9903
0.0112	116.0	1160	0.0427	0.8459	0.8942	0.9901	nan	0.7924	0.9961	nan	0.7020	0.9898
0.0132	118.0	1180	0.0407	0.8510	0.9011	0.9904	nan	0.8062	0.9960	nan	0.7118	0.9901
0.0084	120.0	1200	0.0432	0.8510	0.9015	0.9903	nan	0.8071	0.9959	nan	0.7118	0.9901
0.008	122.0	1220	0.0431	0.8504	0.9077	0.9901	nan	0.8202	0.9953	nan	0.7109	0.9899
0.0069	124.0	1240	0.0424	0.8522	0.8982	0.9905	nan	0.8001	0.9963	nan	0.7141	0.9903
0.006	126.0	1260	0.0447	0.8537	0.9114	0.9904	nan	0.8275	0.9953	nan	0.7173	0.9901
0.0123	128.0	1280	0.0464	0.8529	0.9102	0.9903	nan	0.8250	0.9954	nan	0.7157	0.9901
0.0073	130.0	1300	0.0441	0.8520	0.9025	0.9904	nan	0.8090	0.9959	nan	0.7139	0.9902
0.0066	132.0	1320	0.0447	0.8524	0.9086	0.9903	nan	0.8217	0.9954	nan	0.7148	0.9901
0.0063	134.0	1340	0.0434	0.8546	0.9077	0.9905	nan	0.8197	0.9957	nan	0.7189	0.9903
0.0068	136.0	1360	0.0475	0.8518	0.9090	0.9902	nan	0.8226	0.9953	nan	0.7135	0.9900
0.0056	138.0	1380	0.0458	0.8549	0.9122	0.9905	nan	0.8291	0.9954	nan	0.7195	0.9902
0.007	140.0	1400	0.0455	0.8554	0.9126	0.9905	nan	0.8298	0.9954	nan	0.7205	0.9903
0.0064	142.0	1420	0.0476	0.8542	0.9047	0.9906	nan	0.8133	0.9960	nan	0.7180	0.9903
0.0065	144.0	1440	0.0437	0.8556	0.9107	0.9906	nan	0.8258	0.9956	nan	0.7210	0.9903
0.005	146.0	1460	0.0455	0.8551	0.9098	0.9905	nan	0.8239	0.9956	nan	0.7198	0.9903
0.005	148.0	1480	0.0458	0.8539	0.9084	0.9905	nan	0.8212	0.9956	nan	0.7175	0.9902
0.0048	150.0	1500	0.0462	0.8558	0.9041	0.9907	nan	0.8121	0.9962	nan	0.7211	0.9905
0.0063	152.0	1520	0.0453	0.8560	0.9175	0.9904	nan	0.8400	0.9950	nan	0.7217	0.9902
0.006	154.0	1540	0.0473	0.8531	0.9073	0.9904	nan	0.8190	0.9956	nan	0.7160	0.9902
0.0043	156.0	1560	0.0448	0.8562	0.9100	0.9906	nan	0.8243	0.9957	nan	0.7220	0.9904
0.0049	158.0	1580	0.0480	0.8518	0.9137	0.9901	nan	0.8324	0.9949	nan	0.7138	0.9899
0.0065	160.0	1600	0.0475	0.8556	0.9095	0.9906	nan	0.8233	0.9957	nan	0.7209	0.9903
0.0052	162.0	1620	0.0479	0.8531	0.9087	0.9904	nan	0.8218	0.9955	nan	0.7161	0.9901
0.0063	164.0	1640	0.0488	0.8571	0.9115	0.9907	nan	0.8273	0.9956	nan	0.7238	0.9904
0.0053	166.0	1660	0.0514	0.8515	0.9152	0.9901	nan	0.8357	0.9948	nan	0.7132	0.9898
0.0046	168.0	1680	0.0476	0.8540	0.9040	0.9906	nan	0.8119	0.9960	nan	0.7177	0.9903
0.0039	170.0	1700	0.0483	0.5699	0.9121	0.9905	nan	0.8289	0.9954	0.0	0.7195	0.9902
0.0044	172.0	1720	0.0494	0.8550	0.9114	0.9905	nan	0.8273	0.9954	nan	0.7197	0.9902
0.0051	174.0	1740	0.0503	0.8556	0.9103	0.9906	nan	0.8250	0.9956	nan	0.7208	0.9903
0.0041	176.0	1760	0.0499	0.8545	0.9118	0.9904	nan	0.8283	0.9954	nan	0.7188	0.9902
0.0049	178.0	1780	0.0525	0.8541	0.9066	0.9905	nan	0.8174	0.9958	nan	0.7179	0.9903
0.0048	180.0	1800	0.0496	0.8556	0.9165	0.9904	nan	0.8380	0.9951	nan	0.7210	0.9902
0.008	182.0	1820	0.0487	0.8528	0.9085	0.9904	nan	0.8215	0.9955	nan	0.7155	0.9901
0.0041	184.0	1840	0.0506	0.8519	0.9125	0.9902	nan	0.8300	0.9950	nan	0.7138	0.9899

Framework versions

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