Problems of traing

[jaceqin]

I encountered the following problems when training with BRATS dataset！Can you help me？Thanks！

File "D:\jace\pythonProject\MedSegDiff-master\MedSegDiff-master\guided_diffusion\train_util.py", line 83, in init self._load_and_sync_parameters() File "D:\jace\pythonProject\MedSegDiff-master\MedSegDiff-master\guided_diffusion\train_util.py", line 139, in _load_and_sync_parameters dist_util.sync_params(self.model.parameters()) File "D:\jace\pythonProject\MedSegDiff-master\MedSegDiff-master\guided_diffusion\dist_util.py", line 76, in sync_params dist.broadcast(p, 0) RuntimeError: a leaf Variable that requires grad is being used in an in-place operation.

[WuJunde]

have you modified anything in the code?

[litlepeng]

I directly trained on the BraTS20 dataset and reported the following error. Please help me answer it. Thank you.

[WuJunde]

It processes each slice individually, you should input a slice with the shape [8,5,256,256].

[litlepeng]

I directly used the BraTS20 dataset for training, now how do I need to modify it specifically, thank you

[WuJunde]

you need to extract each slice from a volume for training. you can also download brats20 dataset with slice format from kaggle

[litlepeng]

Would it be convenient for you to put a link? I also could not find the sliced brats20 dataset on kaggle. Thank you very much

[WuJunde]

i see it provides slices: https://www.kaggle.com/awsaf49/brats2020-training-data. but i think extract slices is simpler

[litlepeng]

Thank you. I'll try again.

[litlepeng]

The brats20 of this slice is in h5 format, and it is not consistent with the data sets t1, t1ce, t2, flair that you need. Now do you need to change the bratsloard.py file name yourself? Could you please provide your dataset directly? Thank you very much。

[WuJunde]

hi, in fact, you can simply extract the slices from volume through two lines code like: for index in range(volume.size(2)): slice = volume[:,:,index,:,:] ps: MedSegDiff is not a good project for you to start in this field, a recomment route is unet->nnunet->transunet->medsegdiff.

[litlepeng]

Thank you for your advice