mit-b0-03-04-25-15-21_necrosis

This model is a fine-tuned version of nvidia/mit-b0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1459
• Mean Iou: 0.7491
• Mean Accuracy: 0.8053
• Overall Accuracy: 0.9465
• Accuracy Background: 0.9950
• Accuracy Necrosis: 0.5999
• Accuracy Root: 0.8210
• Iou Background: 0.9514
• Iou Necrosis: 0.5271
• Iou Root: 0.7687

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: 6e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Necrosis	Accuracy Root	Iou Background	Iou Necrosis	Iou Root
1.0286	0.625	20	1.0682	0.2971	0.5482	0.4605	0.3950	0.5573	0.6921	0.3931	0.0404	0.4577
0.8087	1.25	40	0.9066	0.4491	0.6758	0.6734	0.6231	0.5144	0.8901	0.6221	0.0597	0.6654
0.5207	1.875	60	0.7452	0.5011	0.7573	0.7185	0.6773	0.7209	0.8737	0.6762	0.0877	0.7393
0.4503	2.5	80	0.5908	0.5902	0.8478	0.8390	0.8422	0.8815	0.8199	0.8346	0.1719	0.7639
0.3617	3.125	100	0.7355	0.4588	0.7862	0.7203	0.6706	0.7915	0.8965	0.6659	0.1379	0.5726
0.2934	3.75	120	0.3994	0.6875	0.8473	0.9220	0.9489	0.7428	0.8502	0.9244	0.3873	0.7507
0.3384	4.375	140	0.3423	0.6961	0.8540	0.9276	0.9587	0.7662	0.8371	0.9332	0.3891	0.7659
0.3014	5.0	160	0.3397	0.6933	0.8418	0.9209	0.9532	0.7439	0.8282	0.9239	0.4393	0.7167
0.3316	5.625	180	0.3360	0.6717	0.8395	0.9088	0.9339	0.7433	0.8414	0.9076	0.4098	0.6979
0.2285	6.25	200	0.2552	0.7329	0.8140	0.9439	0.9727	0.5721	0.8973	0.9498	0.4681	0.7810
0.1486	6.875	220	0.2143	0.7410	0.8183	0.9470	0.9837	0.6061	0.8651	0.9526	0.4807	0.7898
0.2264	7.5	240	0.2154	0.7168	0.7719	0.9352	0.9956	0.5491	0.7710	0.9366	0.4871	0.7265
0.172	8.125	260	0.1926	0.7330	0.7920	0.9444	0.9906	0.5496	0.8358	0.9491	0.4799	0.7700
0.1841	8.75	280	0.2086	0.7252	0.8199	0.9352	0.9926	0.7112	0.7559	0.9405	0.5125	0.7225
0.1422	9.375	300	0.1668	0.7427	0.8002	0.9473	0.9922	0.5676	0.8409	0.9528	0.4972	0.7781
0.1747	10.0	320	0.1877	0.7257	0.8219	0.9401	0.9943	0.6956	0.7756	0.9496	0.4864	0.7409
0.1492	10.625	340	0.2011	0.7252	0.8346	0.9387	0.9887	0.7309	0.7842	0.9485	0.4869	0.7402
0.1835	11.25	360	0.1758	0.7429	0.8284	0.9438	0.9953	0.7006	0.7893	0.9512	0.5228	0.7547
0.2357	11.875	380	0.1770	0.7439	0.8264	0.9441	0.9941	0.6874	0.7978	0.9508	0.5223	0.7587
0.1232	12.5	400	0.1619	0.7534	0.8204	0.9504	0.9918	0.6205	0.8490	0.9575	0.5127	0.7901
0.1584	13.125	420	0.1911	0.7264	0.8091	0.9366	0.9947	0.6714	0.7613	0.9416	0.5111	0.7265
0.2099	13.75	440	0.1593	0.7495	0.8354	0.9485	0.9868	0.6695	0.8500	0.9563	0.5053	0.7869
0.2723	14.375	460	0.1668	0.7353	0.8029	0.9426	0.9906	0.5995	0.8185	0.9477	0.4999	0.7583
0.1406	15.0	480	0.1714	0.7404	0.8172	0.9426	0.9918	0.6555	0.8043	0.9477	0.5170	0.7564
0.1624	15.625	500	0.1527	0.7445	0.8007	0.9511	0.9913	0.5435	0.8674	0.9591	0.4811	0.7934
0.1519	16.25	520	0.1503	0.7459	0.8044	0.9495	0.9936	0.5743	0.8455	0.9568	0.4973	0.7837
0.0996	16.875	540	0.1437	0.7551	0.8216	0.9524	0.9898	0.6065	0.8685	0.9607	0.5066	0.7981
0.3985	17.5	560	0.1520	0.7355	0.7891	0.9472	0.9931	0.5310	0.8431	0.9535	0.4759	0.7771
0.0979	18.125	580	0.1586	0.7447	0.8072	0.9453	0.9940	0.6105	0.8171	0.9502	0.5178	0.7661
0.1027	18.75	600	0.1482	0.7551	0.8261	0.9482	0.9937	0.6605	0.8242	0.9544	0.5341	0.7768
0.1372	19.375	620	0.1504	0.7487	0.8107	0.9463	0.9947	0.6195	0.8180	0.9512	0.5256	0.7693
0.1219	20.0	640	0.1473	0.7558	0.8297	0.9488	0.9944	0.6723	0.8223	0.9556	0.5331	0.7786
0.1045	20.625	660	0.1827	0.7228	0.8050	0.9333	0.9962	0.6811	0.7375	0.9373	0.5231	0.7080
0.1034	21.25	680	0.1534	0.7489	0.8431	0.9451	0.9939	0.7407	0.7946	0.9537	0.5347	0.7583
0.2095	21.875	700	0.1469	0.7401	0.7944	0.9477	0.9935	0.5486	0.8411	0.9543	0.4900	0.7760
0.2314	22.5	720	0.1474	0.7529	0.8222	0.9493	0.9933	0.6387	0.8345	0.9566	0.5206	0.7814
0.1133	23.125	740	0.1645	0.7352	0.8301	0.9404	0.9956	0.7276	0.7671	0.9486	0.5204	0.7366
0.0817	23.75	760	0.1341	0.7563	0.8089	0.9537	0.9936	0.5653	0.8678	0.9612	0.5056	0.8020
0.1422	24.375	780	0.1359	0.7622	0.8287	0.9521	0.9919	0.6404	0.8537	0.9595	0.5333	0.7937
0.0963	25.0	800	0.1450	0.7558	0.8365	0.9487	0.9945	0.6977	0.8172	0.9568	0.5359	0.7748
0.085	25.625	820	0.1543	0.7370	0.7929	0.9441	0.9952	0.5704	0.8132	0.9491	0.5024	0.7596
0.2654	26.25	840	0.1599	0.7338	0.7903	0.9424	0.9957	0.5727	0.8027	0.9472	0.5030	0.7511
0.1506	26.875	860	0.1503	0.7478	0.8091	0.9466	0.9940	0.6096	0.8236	0.9526	0.5227	0.7681
0.114	27.5	880	0.1440	0.7463	0.8029	0.9488	0.9938	0.5738	0.8412	0.9556	0.5037	0.7795
0.0748	28.125	900	0.1611	0.7362	0.7949	0.9416	0.9956	0.5935	0.7957	0.9461	0.5162	0.7463
0.1368	28.75	920	0.1457	0.7496	0.8074	0.9477	0.9948	0.5993	0.8281	0.9535	0.5214	0.7737
0.0836	29.375	940	0.1622	0.7170	0.7663	0.9408	0.9951	0.4931	0.8106	0.9464	0.4579	0.7469
0.215	30.0	960	0.1340	0.7450	0.7959	0.9511	0.9945	0.5366	0.8566	0.9588	0.4857	0.7905
0.1397	30.625	980	0.1370	0.7417	0.7931	0.9495	0.9945	0.5367	0.8480	0.9568	0.4853	0.7830
0.1468	31.25	1000	0.1373	0.7591	0.8221	0.9507	0.9944	0.6332	0.8387	0.9574	0.5337	0.7863
0.0733	31.875	1020	0.1380	0.7439	0.7946	0.9500	0.9943	0.5380	0.8515	0.9573	0.4894	0.7851
0.1454	32.5	1040	0.1414	0.7522	0.8106	0.9487	0.9947	0.6048	0.8324	0.9548	0.5235	0.7781
0.1203	33.125	1060	0.1459	0.7498	0.8197	0.9467	0.9952	0.6512	0.8126	0.9531	0.5282	0.7681
0.2697	33.75	1080	0.1381	0.7541	0.8128	0.9494	0.9944	0.6080	0.8359	0.9554	0.5256	0.7815
0.0884	34.375	1100	0.1629	0.7276	0.7795	0.9403	0.9961	0.5474	0.7950	0.9443	0.4961	0.7426
0.1911	35.0	1120	0.1395	0.7585	0.8293	0.9501	0.9935	0.6603	0.8341	0.9574	0.5354	0.7827
0.129	35.625	1140	0.1709	0.7278	0.7804	0.9399	0.9955	0.5512	0.7946	0.9440	0.4991	0.7402
0.0965	36.25	1160	0.1409	0.7472	0.7995	0.9490	0.9948	0.5633	0.8403	0.9552	0.5057	0.7806
0.0956	36.875	1180	0.1403	0.7389	0.7885	0.9489	0.9942	0.5223	0.8491	0.9559	0.4793	0.7817
0.1023	37.5	1200	0.1512	0.7438	0.8022	0.9447	0.9952	0.6001	0.8113	0.9495	0.5208	0.7610
0.1341	38.125	1220	0.1527	0.7422	0.7961	0.9450	0.9953	0.5771	0.8159	0.9496	0.5143	0.7627
0.0669	38.75	1240	0.1340	0.7459	0.7950	0.9516	0.9938	0.5282	0.8630	0.9591	0.4851	0.7933
0.2038	39.375	1260	0.1347	0.7572	0.8120	0.9509	0.9946	0.5967	0.8446	0.9568	0.5269	0.7879
0.1287	40.0	1280	0.1554	0.7415	0.7959	0.9433	0.9957	0.5872	0.8046	0.9471	0.5224	0.7550
0.1505	40.625	1300	0.1353	0.7470	0.7972	0.9499	0.9946	0.5491	0.8477	0.9563	0.4997	0.7851
0.0827	41.25	1320	0.1408	0.7522	0.8089	0.9481	0.9947	0.6021	0.8300	0.9535	0.5274	0.7757
0.1537	41.875	1340	0.1469	0.7468	0.8007	0.9458	0.9950	0.5874	0.8196	0.9504	0.5244	0.7656
0.1328	42.5	1360	0.1415	0.7490	0.8030	0.9477	0.9948	0.5835	0.8306	0.9531	0.5199	0.7741
0.0971	43.125	1380	0.1330	0.7578	0.8133	0.9515	0.9943	0.5966	0.8489	0.9580	0.5252	0.7903
0.1021	43.75	1400	0.1332	0.7520	0.8033	0.9512	0.9945	0.5634	0.8521	0.9579	0.5086	0.7894
0.0707	44.375	1420	0.1404	0.7496	0.8029	0.9481	0.9952	0.5825	0.8311	0.9536	0.5201	0.7753
0.0767	45.0	1440	0.1388	0.7520	0.8066	0.9486	0.9949	0.5916	0.8332	0.9543	0.5243	0.7775
0.0747	45.625	1460	0.1351	0.7594	0.8200	0.9504	0.9944	0.6274	0.8383	0.9567	0.5369	0.7846
0.2155	46.25	1480	0.1413	0.7549	0.8147	0.9478	0.9949	0.6254	0.8237	0.9531	0.5384	0.7732
0.0757	46.875	1500	0.1379	0.7560	0.8147	0.9495	0.9944	0.6137	0.8359	0.9555	0.5314	0.7810
0.1457	47.5	1520	0.1528	0.7459	0.8057	0.9441	0.9955	0.6174	0.8042	0.9484	0.5323	0.7570
0.0952	48.125	1540	0.1542	0.7467	0.8072	0.9438	0.9955	0.6246	0.8015	0.9479	0.5368	0.7556
0.1606	48.75	1560	0.1465	0.7526	0.8136	0.9464	0.9950	0.6303	0.8154	0.9512	0.5393	0.7672
0.1153	49.375	1580	0.1411	0.7511	0.8063	0.9483	0.9946	0.5916	0.8328	0.9539	0.5225	0.7768
0.065	50.0	1600	0.1459	0.7491	0.8053	0.9465	0.9950	0.5999	0.8210	0.9514	0.5271	0.7687

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0