segformer-b5-finetuned-segments-metalchip-4-Mar-10epochs

This model is a fine-tuned version of nvidia/mit-b5 on the segments/metalchip-semantic dataset. It achieves the following results on the evaluation set:

• Loss: 0.1278
• Mean Iou: 0.9007
• Mean Accuracy: 0.9476
• Overall Accuracy: 0.9478
• Accuracy Background: 0.9538
• Accuracy Metal lines: 0.9414
• Iou Background: 0.9047
• Iou Metal lines: 0.8966

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: 0.01
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Metal lines	Iou Background	Iou Metal lines
0.2928	0.25	20	0.5604	0.7372	0.8474	0.8531	0.9962	0.6985	0.7788	0.6957
0.234	0.51	40	0.4095	0.5873	0.7566	0.7479	0.5316	0.9815	0.5227	0.6518
0.2606	0.76	60	0.1990	0.8682	0.9282	0.9298	0.9703	0.8861	0.8777	0.8586
0.1575	1.01	80	0.1853	0.8604	0.9238	0.9254	0.9670	0.8806	0.8706	0.8502
0.155	1.27	100	0.3476	0.7775	0.8728	0.8775	0.9935	0.7522	0.8080	0.7469
0.1565	1.52	120	0.1813	0.8615	0.9243	0.9260	0.9689	0.8798	0.8718	0.8512
0.1773	1.77	140	0.1656	0.8830	0.9382	0.9379	0.9297	0.9467	0.8860	0.8799
0.1407	2.03	160	0.1653	0.8841	0.9378	0.9387	0.9629	0.9126	0.8908	0.8775
0.1692	2.28	180	0.1726	0.8720	0.9331	0.9316	0.8949	0.9712	0.8717	0.8723
0.1525	2.53	200	0.1415	0.8924	0.9434	0.9432	0.9383	0.9484	0.8955	0.8892
0.1434	2.78	220	0.3651	0.7260	0.8487	0.8431	0.7012	0.9963	0.6988	0.7533
0.1701	3.04	240	0.1505	0.8848	0.9396	0.9389	0.9216	0.9576	0.8868	0.8829
0.093	3.29	260	0.1391	0.8931	0.9438	0.9436	0.9370	0.9507	0.8961	0.8901
0.1331	3.54	280	0.1363	0.8936	0.9440	0.9439	0.9395	0.9486	0.8968	0.8904
0.124	3.8	300	0.1329	0.8960	0.9451	0.9452	0.9484	0.9418	0.8999	0.8921
0.144	4.05	320	0.1461	0.8847	0.9381	0.9391	0.9642	0.9120	0.8915	0.8780
0.1226	4.3	340	0.1463	0.8884	0.9416	0.9409	0.9239	0.9592	0.8903	0.8864
0.1344	4.56	360	0.1323	0.8986	0.9467	0.9467	0.9453	0.9482	0.9020	0.8953
0.1497	4.81	380	0.1382	0.8929	0.9435	0.9435	0.9418	0.9453	0.8964	0.8894
0.146	5.06	400	0.1382	0.8938	0.9434	0.9441	0.9600	0.9268	0.8991	0.8885
0.1046	5.32	420	0.1449	0.8899	0.9419	0.9418	0.9401	0.9436	0.8934	0.8863
0.1355	5.57	440	0.1378	0.8949	0.9448	0.9446	0.9386	0.9511	0.8979	0.8919
0.1268	5.82	460	0.1334	0.8976	0.9461	0.9461	0.9449	0.9473	0.9010	0.8941
0.1146	6.08	480	0.1312	0.8990	0.9468	0.9469	0.9500	0.9435	0.9028	0.8952
0.13	6.33	500	0.1315	0.9000	0.9474	0.9475	0.9490	0.9457	0.9036	0.8964
0.1196	6.58	520	0.1325	0.8981	0.9466	0.9464	0.9401	0.9531	0.9010	0.8952
0.1111	6.84	540	0.1298	0.8987	0.9465	0.9467	0.9507	0.9424	0.9025	0.8948
0.095	7.09	560	0.1351	0.8990	0.9465	0.9469	0.9593	0.9336	0.9037	0.8943
0.1077	7.34	580	0.1316	0.8990	0.9467	0.9469	0.9530	0.9404	0.9031	0.8950
0.1246	7.59	600	0.1416	0.8896	0.9414	0.9417	0.9502	0.9325	0.8943	0.8849
0.1108	7.85	620	0.1406	0.8907	0.9413	0.9424	0.9711	0.9115	0.8975	0.8839
0.148	8.1	640	0.1337	0.8949	0.9441	0.9447	0.9591	0.9291	0.9000	0.8898
0.1056	8.35	660	0.1290	0.9007	0.9474	0.9478	0.9581	0.9368	0.9051	0.8962
0.1003	8.61	680	0.1291	0.8972	0.9458	0.9459	0.9486	0.9429	0.9010	0.8934
0.1089	8.86	700	0.1295	0.9008	0.9476	0.9479	0.9560	0.9392	0.9050	0.8966
0.0924	9.11	720	0.1297	0.9004	0.9476	0.9477	0.9500	0.9451	0.9041	0.8967
0.0851	9.37	740	0.1406	0.8937	0.9430	0.9440	0.9694	0.9166	0.9000	0.8873
0.1223	9.62	760	0.1341	0.8983	0.9461	0.9466	0.9593	0.9329	0.9031	0.8936
0.1592	9.87	780	0.1306	0.8985	0.9464	0.9466	0.9522	0.9406	0.9025	0.8944
0.1344	10.13	800	0.1293	0.8991	0.9469	0.9470	0.9483	0.9455	0.9028	0.8955
0.1216	10.38	820	0.1309	0.8982	0.9463	0.9465	0.9521	0.9404	0.9023	0.8942
0.1086	10.63	840	0.1327	0.8981	0.9462	0.9464	0.9531	0.9393	0.9023	0.8940
0.1199	10.89	860	0.1323	0.8990	0.9465	0.9469	0.9580	0.9350	0.9036	0.8944
0.1148	11.14	880	0.1290	0.9005	0.9477	0.9477	0.9479	0.9475	0.9040	0.8970
0.1368	11.39	900	0.1298	0.9000	0.9471	0.9475	0.9571	0.9370	0.9044	0.8955
0.1289	11.65	920	0.1335	0.8952	0.9447	0.9447	0.9448	0.9447	0.8988	0.8916
0.1181	11.9	940	0.1305	0.8997	0.9471	0.9473	0.9524	0.9418	0.9037	0.8957
0.1738	12.15	960	0.1278	0.9013	0.9480	0.9482	0.9519	0.9441	0.9051	0.8976
0.1269	12.41	980	0.1285	0.9000	0.9472	0.9475	0.9546	0.9398	0.9042	0.8959
0.1118	12.66	1000	0.1323	0.8967	0.9450	0.9457	0.9636	0.9263	0.9021	0.8913
0.0688	12.91	1020	0.1302	0.8998	0.9472	0.9473	0.9503	0.9441	0.9035	0.8960
0.1245	13.16	1040	0.1295	0.9002	0.9474	0.9476	0.9524	0.9423	0.9041	0.8963
0.0838	13.42	1060	0.1303	0.9007	0.9474	0.9479	0.9603	0.9345	0.9054	0.8961
0.1103	13.67	1080	0.1285	0.8994	0.9469	0.9471	0.9519	0.9419	0.9033	0.8954
0.1204	13.92	1100	0.1326	0.8988	0.9468	0.9468	0.9463	0.9474	0.9023	0.8954
0.0797	14.18	1120	0.1301	0.8989	0.9462	0.9469	0.9634	0.9291	0.9040	0.8937
0.1192	14.43	1140	0.1283	0.8998	0.9470	0.9474	0.9560	0.9381	0.9042	0.8955
0.1713	14.68	1160	0.1282	0.9015	0.9479	0.9483	0.9577	0.9381	0.9058	0.8971
0.1032	14.94	1180	0.1340	0.8957	0.9445	0.9451	0.9624	0.9265	0.9011	0.8904
0.1097	15.19	1200	0.1300	0.9003	0.9471	0.9476	0.9613	0.9329	0.9051	0.8955
0.194	15.44	1220	0.1269	0.9014	0.9479	0.9482	0.9568	0.9389	0.9056	0.8971
0.0849	15.7	1240	0.1288	0.8997	0.9468	0.9473	0.9609	0.9327	0.9045	0.8949
0.105	15.95	1260	0.1293	0.9003	0.9472	0.9476	0.9590	0.9353	0.9048	0.8957
0.1102	16.2	1280	0.1268	0.9008	0.9478	0.9479	0.9513	0.9442	0.9046	0.8971
0.1185	16.46	1300	0.1301	0.8998	0.9471	0.9473	0.9546	0.9395	0.9040	0.8956
0.1087	16.71	1320	0.1280	0.9001	0.9473	0.9475	0.9512	0.9435	0.9039	0.8963
0.1033	16.96	1340	0.1281	0.9002	0.9474	0.9475	0.9501	0.9447	0.9039	0.8965
0.1083	17.22	1360	0.1270	0.9010	0.9477	0.9480	0.9549	0.9406	0.9051	0.8969
0.124	17.47	1380	0.1285	0.9007	0.9476	0.9478	0.9536	0.9416	0.9047	0.8967
0.0835	17.72	1400	0.1288	0.8996	0.9468	0.9472	0.9583	0.9353	0.9041	0.8950
0.1263	17.97	1420	0.1274	0.9012	0.9477	0.9482	0.9584	0.9371	0.9056	0.8968
0.1073	18.23	1440	0.1274	0.9008	0.9476	0.9479	0.9562	0.9389	0.9050	0.8965
0.1323	18.48	1460	0.1282	0.9001	0.9475	0.9475	0.9488	0.9461	0.9037	0.8965
0.0845	18.73	1480	0.1272	0.9007	0.9475	0.9479	0.9556	0.9394	0.9049	0.8965
0.1131	18.99	1500	0.1279	0.9008	0.9476	0.9479	0.9545	0.9407	0.9048	0.8967
0.1198	19.24	1520	0.1306	0.8987	0.9466	0.9467	0.9492	0.9441	0.9024	0.8950
0.119	19.49	1540	0.1275	0.9009	0.9476	0.9480	0.9565	0.9388	0.9052	0.8966
0.0984	19.75	1560	0.1275	0.9007	0.9477	0.9478	0.9519	0.9434	0.9045	0.8969
0.0955	20.0	1580	0.1278	0.9007	0.9476	0.9478	0.9538	0.9414	0.9047	0.8966

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2