Retrain results do not match the paper result.

[anothersin]

Thank you very much for the very good code. But some parts still confuse me. We have retrained the new model following the steps on Readme, but either VT-UNET-S or VT-UNET-B is somewhat different from Table 1 in the paper (e.g. VT-UNET-B, retrained: Dice_et0.80674840333256,Dice_ tc0.8525805936607558, Dice_wt0.9158492648036342 hd_et4.286296144927589,hd_tc4.707234057781734,hd_wt3.6658415312320956 dice0.8583927539323167 HD4.219790577980473 ). We speculate whether this is caused by different test sets. We are dividing the test set according to VT-UNet/VTUNet/dataset.json, is it correct? If it is convenient, could you please provide the VT-UNET-B model that has been trained.