Pyfftw and scikit-image python package conflict

Install, Configure and Update
1

Hi,

I am attempting to install cryosparc using the link + licence provided in the email and am running into the following unresolvable python package dependencies.

UnsatisfiableError: The following specifications were found to be in conflict:

  • pyfftw=0.9.2
  • scikit-image=0.12.3

My system is running Linux Mint (uname -a | Linux 4.10.0-42-generic #46~16.04.1-Ubuntu SMP), with nvidia 387.26 drivers (nvidia-smi | NVIDIA-SMI 387.26 Driver Version: 387.26) and CUDA 9 (nvcc --version | cuda compilation tools, release 9.1, V9.1.85).

I have chosen to install anaconda directly into the cryosparc directory as suggested by the installer.

Any information or guidance would be greatly appreciated.

Thanks!
joey

2

Hi Joey,

I’m running similar configuration, but with Ubuntu 16.04 and I had the same problem. I’m using cuda 8.0 instead of 9 (as suggested by the developer)

In the end, I fixed it by using another pyfftw package and it seems to run smoothly.

To do so replace the line :

conda install -q -y --force -c richli fftw=3.3.4 pyfftw=0.9.2

by the two lines :

conda install -q -y --force -c richli fftw=3.3.4 
conda install -q -y --force -c conda-forge pyfftw

Note that it’s probably only a temporary fix and a proper installation should probably be done once the problem is fixed.

Hope this help.

Alain

3

Hi Alain,

Something similar ended up working for us as well, and, as you mention, everything seems to be running smoothly so far.

Thanks!
joey

4

This issue is now fixed in v0.6.5
We did end up switching to conda-forge because the other channel no longer maintains pyfftw binaries.

5