safety-utcustom-train-SF30-RGB-b5

This model is a fine-tuned version of nvidia/mit-b5 on the sam1120/safety-utcustom-TRAIN-30 dataset. It achieves the following results on the evaluation set:

• Accuracy Safe: 0.8299
• Accuracy Unlabeled: nan
• Accuracy Unsafe: 0.9036
• Iou Safe: 0.3480
• Iou Unlabeled: 0.0
• Iou Unsafe: 0.8996
• Loss: 0.5783
• Mean Accuracy: 0.8668
• Mean Iou: 0.4158
• Overall Accuracy: 0.9013

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: 2e-06
• 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: 120

Training results

Training Loss	Epoch	Step	Accuracy Safe	Accuracy Unlabeled	Accuracy Unsafe	Iou Safe	Iou Unlabeled	Iou Unsafe	Validation Loss	Mean Accuracy	Mean Iou	Overall Accuracy
1.0614	5.0	10	0.1904	nan	0.5439	0.0682	0.0	0.5350	1.0385	0.3672	0.2011	0.5327
1.0269	10.0	20	0.4801	nan	0.5773	0.1795	0.0	0.5719	0.9975	0.5287	0.2505	0.5742
1.0005	15.0	30	0.6270	nan	0.6316	0.2261	0.0	0.6269	0.9428	0.6293	0.2843	0.6315
0.9716	20.0	40	0.6870	nan	0.6802	0.2529	0.0	0.6756	0.8918	0.6836	0.3095	0.6804
0.9255	25.0	50	0.7339	nan	0.7081	0.2805	0.0	0.7037	0.8542	0.7210	0.3281	0.7089
0.9256	30.0	60	0.7705	nan	0.7229	0.2781	0.0	0.7189	0.8330	0.7467	0.3324	0.7244
0.8167	35.0	70	0.7622	nan	0.7349	0.3004	0.0	0.7311	0.8114	0.7485	0.3438	0.7358
0.7927	40.0	80	0.7776	nan	0.7594	0.3154	0.0	0.7559	0.7793	0.7685	0.3571	0.7600
0.8227	45.0	90	0.8020	nan	0.7821	0.3152	0.0	0.7789	0.7574	0.7920	0.3647	0.7827
0.81	50.0	100	0.8114	nan	0.7983	0.3140	0.0	0.7955	0.7370	0.8049	0.3698	0.7987
0.7198	55.0	110	0.8002	nan	0.8194	0.3303	0.0	0.8162	0.7118	0.8098	0.3822	0.8188
0.7523	60.0	120	0.7877	nan	0.8482	0.3457	0.0	0.8443	0.6832	0.8179	0.3967	0.8462
0.7239	65.0	130	0.8112	nan	0.8485	0.3197	0.0	0.8453	0.6745	0.8298	0.3883	0.8473
0.6235	70.0	140	0.7906	nan	0.8686	0.3507	0.0	0.8649	0.6419	0.8296	0.4052	0.8662
0.6887	75.0	150	0.7951	nan	0.8758	0.3568	0.0	0.8720	0.6302	0.8354	0.4096	0.8732
0.6079	80.0	160	0.8069	nan	0.8879	0.3561	0.0	0.8841	0.6120	0.8474	0.4134	0.8853
0.6022	85.0	170	0.8126	nan	0.9062	0.3699	0.0	0.9020	0.5849	0.8594	0.4240	0.9032
0.5748	90.0	180	0.8053	nan	0.9047	0.3793	0.0	0.9005	0.5802	0.8550	0.4266	0.9016
0.6228	95.0	190	0.8164	nan	0.9050	0.3624	0.0	0.9007	0.5793	0.8607	0.4210	0.9022
0.5332	100.0	200	0.8214	nan	0.9134	0.3623	0.0	0.9091	0.5616	0.8674	0.4238	0.9105
0.6655	105.0	210	0.8262	nan	0.9072	0.3572	0.0	0.9031	0.5688	0.8667	0.4201	0.9046
0.5835	110.0	220	0.8233	nan	0.9092	0.3599	0.0	0.9050	0.5653	0.8662	0.4216	0.9064
0.5764	115.0	230	0.8099	nan	0.9165	0.3783	0.0	0.9120	0.5460	0.8632	0.4301	0.9131
0.5621	120.0	240	0.8299	nan	0.9036	0.3480	0.0	0.8996	0.5783	0.8668	0.4158	0.9013

Framework versions

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