safety-utcustom-train-SF30-RGBD-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:

• Loss: 0.1952
• Mean Iou: 0.6486
• Mean Accuracy: 0.7199
• Overall Accuracy: 0.9704
• Accuracy Unlabeled: nan
• Accuracy Safe: 0.4523
• Accuracy Unsafe: 0.9874
• Iou Unlabeled: nan
• Iou Safe: 0.3271
• Iou Unsafe: 0.9700

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: 1e-05
• 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: 100

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
0.8758	5.0	10	0.9831	0.3415	0.6100	0.9154	nan	0.2839	0.9362	0.0	0.1099	0.9147
0.7637	10.0	20	0.7236	0.3771	0.6275	0.9582	nan	0.2745	0.9806	0.0	0.1735	0.9578
0.6698	15.0	30	0.5510	0.3789	0.6286	0.9593	nan	0.2755	0.9818	0.0	0.1776	0.9590
0.5935	20.0	40	0.4632	0.3822	0.6388	0.9591	nan	0.2967	0.9809	0.0	0.1877	0.9588
0.5108	25.0	50	0.4239	0.3814	0.6492	0.9560	nan	0.3214	0.9769	0.0	0.1887	0.9556
0.4597	30.0	60	0.4134	0.3845	0.6422	0.9596	nan	0.3034	0.9811	0.0	0.1943	0.9592
0.4307	35.0	70	0.3918	0.3900	0.6516	0.9594	nan	0.3229	0.9803	0.0	0.2111	0.9590
0.367	40.0	80	0.3578	0.3885	0.6600	0.9582	nan	0.3415	0.9784	0.0	0.2077	0.9577
0.3249	45.0	90	0.3395	0.3921	0.6587	0.9607	nan	0.3360	0.9813	0.0	0.2161	0.9603
0.292	50.0	100	0.3124	0.3969	0.6622	0.9633	nan	0.3408	0.9837	0.0	0.2280	0.9629
0.2766	55.0	110	0.2820	0.4078	0.6878	0.9644	nan	0.3925	0.9831	0.0	0.2594	0.9639
0.2347	60.0	120	0.2673	0.6169	0.7000	0.9641	nan	0.4181	0.9820	nan	0.2701	0.9636
0.226	65.0	130	0.2350	0.6280	0.6854	0.9698	nan	0.3818	0.9891	nan	0.2865	0.9694
0.3262	70.0	140	0.2354	0.6338	0.7125	0.9674	nan	0.4402	0.9848	nan	0.3006	0.9670
0.1991	75.0	150	0.2231	0.6363	0.7169	0.9676	nan	0.4492	0.9846	nan	0.3056	0.9671
0.2106	80.0	160	0.2089	0.6399	0.7152	0.9688	nan	0.4444	0.9860	nan	0.3114	0.9683
0.1995	85.0	170	0.1969	0.6493	0.7179	0.9709	nan	0.4478	0.9880	nan	0.3281	0.9704
0.1981	90.0	180	0.1909	0.6503	0.7136	0.9716	nan	0.4381	0.9892	nan	0.3293	0.9712
0.1875	95.0	190	0.1965	0.6473	0.7231	0.9697	nan	0.4598	0.9864	nan	0.3254	0.9692
0.2088	100.0	200	0.1952	0.6486	0.7199	0.9704	nan	0.4523	0.9874	nan	0.3271	0.9700

Framework versions

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