AttributeError: 'int' object has no attribute 'item'

[facebookresearch]

AttributeError: 'int' object has no attribute 'item'

[btgraham]

@AutumnWormSun Please use PyTorch master: https://github.com/facebookresearch/SparseConvNet#pytorch-setup

[btgraham]

Are you using PyTorch master? (https://github.com/facebookresearch/SparseConvNet#pytorch-setup)
SparseConvNet had to be upgraded recently to be compatible with the latest PyTorch changes.
If you are on master, the following code should work:

import sparseconvnet as scn
scn.SubmanifoldConvolution(3,4,4,3,False).filter_size.prod().item()

If you are already on master, please give a minimal example of code that fails.

[fengxiuyaun]

Thanks! I sloved it. thank you.

[fengxiuyaun]

hi,
Could you share me example of Kaggle Diabetic Retinopathy Detection, 2015?
thanks

[AutumnWormSun]

@fengxiuyaun I encountered the same problem!
Environment: Python 3.6, Pytorch 0.31, SparseConvNet 0.11

Error:
self.filter_volume = self.filter_size.prod().item()
AttributeError: 'int' object has no attribute 'item'

Code:
import sparseconvnet as scn
scn.SubmanifoldConvolution(3,4,4,3,False).filter_size.prod().item()

@btgraham Example and demo you showed both throw exception. There must be some bugs.

[btgraham]

@fengxiuyaun The Diabetes example is described here
https://www.kaggle.com/c/diabetic-retinopathy-detection/discussion/15801
I am not going to import it into this repo as it is not recent research.
Regards
Ben

[AutumnWormSun]

@btgraham Thanks! I will try.