error:
backend = thnn.type2backend[kernel.type()]
—> backend.SpatialGridSamplerBilinear_updateOutput(backend.library_state, kernel, grid, output, 0)
/usr/local/lib/python3.6/dist-packages/torch/_thnn/utils.py
in getattr(self, name)
25 method = self.methods.get(name, None)
26 if method is None:
—>27 raise NotImplementedError
28 return method
NotImplementedError:
I am working on google colab . I am getting this not implemented issue for SpatialGridSamplerBilinear_updateOutput function.Which is present possibly as SpatialGridSamplerBilinear.cu file.
I am working with GPU environment on collab with cuda version : 10.0
!/usr/local/cuda/bin/nvcc --version
nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130
on collab:
pytorch : 1.1.0
torchvision : 0.3.0
python : 3.6.4
Also if use pytorch 0.4.1 then torchvision 0.3.0 is incompatible.
The same code shows no errors in a perfect nvidia environment in a lab (whose access is rare)
and there pytorch is 0.4.1 and in lab there is a sure call to Spatialsamplerbilinear.cu.
I think it is present in THCUNN and in need to somehow incorporate cudnn in collab environment.
can someone help please…