Modelpublic

sam1120/safety-utcustom-train-SF30-RGB-b0

sourceHugging Faceotherupdated 3y agoView on Hugging Face

0likes17downloads

Model Card

safety-utcustom-train-SF30-RGB-b0

This model is a fine-tuned version of nvidia/mit-b0 on the sam1120/safety-utcustom-TRAIN-30 dataset. It achieves the following results on the evaluation set:

—Loss: 0.7492
—Mean Iou: 0.3878
—Mean Accuracy: 0.8431
—Overall Accuracy: 0.9233
—Accuracy Unlabeled: nan
—Accuracy Safe: 0.7575
—Accuracy Unsafe: 0.9287
—Iou Unlabeled: 0.0
—Iou Safe: 0.2418
—Iou Unsafe: 0.9214

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: 9e-06
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_ratio: 0.05
—num_epochs: 120

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Safe	Accuracy Unsafe	Iou Safe	Iou Unsafe
1.1527	5.0	10	1.1085	0.0590	0.4585	0.1664	nan	0.7704	0.1465	0.0307	0.1464
1.1326	10.0	20	1.1091	0.0963	0.6082	0.2699	nan	0.9695	0.2470	0.0419	0.2470
1.0981	15.0	30	1.0980	0.1530	0.6989	0.4242	nan	0.9922	0.4055	0.0535	0.4055
1.086	20.0	40	1.0822	0.1916	0.7515	0.5256	nan	0.9927	0.5103	0.0644	0.5103
1.0466	25.0	50	1.0541	0.2226	0.7909	0.6043	nan	0.9902	0.5917	0.0761	0.5917
1.0533	30.0	60	1.0249	0.2444	0.8167	0.6580	nan	0.9863	0.6472	0.0861	0.6471
0.9779	35.0	70	1.0010	0.2607	0.8322	0.6966	nan	0.9771	0.6874	0.0951	0.6871
0.9161	40.0	80	0.9695	0.2808	0.8487	0.7412	nan	0.9635	0.7339	0.1091	0.7334
0.9843	45.0	90	0.9403	0.3004	0.8631	0.7823	nan	0.9494	0.7768	0.1254	0.7759
0.9568	50.0	100	0.9071	0.3176	0.8663	0.8169	nan	0.9191	0.8135	0.1412	0.8117
0.8443	55.0	110	0.8627	0.3403	0.8656	0.8576	nan	0.8742	0.8570	0.1672	0.8537
0.8765	60.0	120	0.8488	0.3450	0.8625	0.8657	nan	0.8591	0.8659	0.1729	0.8620
0.899	65.0	130	0.8429	0.3481	0.8629	0.8705	nan	0.8548	0.8710	0.1772	0.8669
0.7713	70.0	140	0.8085	0.3632	0.8497	0.8939	nan	0.8026	0.8969	0.1983	0.8912
0.8505	75.0	150	0.7821	0.3762	0.8465	0.9102	nan	0.7786	0.9145	0.2208	0.9079
0.7352	80.0	160	0.7841	0.3819	0.8392	0.9173	nan	0.7557	0.9226	0.2304	0.9153
0.7205	85.0	170	0.7502	0.3974	0.8400	0.9325	nan	0.7413	0.9388	0.2613	0.9309
0.711	90.0	180	0.7417	0.3962	0.8428	0.9313	nan	0.7484	0.9373	0.2591	0.9296
0.7855	95.0	190	0.7281	0.4003	0.8439	0.9343	nan	0.7473	0.9404	0.2683	0.9327
0.7632	100.0	200	0.7494	0.3883	0.8419	0.9237	nan	0.7545	0.9293	0.2430	0.9219
0.8145	105.0	210	0.7495	0.3862	0.8412	0.9219	nan	0.7551	0.9274	0.2387	0.9201
0.8217	110.0	220	0.7355	0.3933	0.8422	0.9282	nan	0.7502	0.9341	0.2533	0.9265
0.7784	115.0	230	0.7258	0.4088	0.8411	0.9413	nan	0.7340	0.9481	0.2864	0.9400
0.8349	120.0	240	0.7492	0.3878	0.8431	0.9233	nan	0.7575	0.9287	0.2418	0.9214

Framework versions

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