fail to create conda environment

[jianhaoc]

Hi, I tried to run init.sh. But it failed to create an conda environment (see below). Did I miss anything? Thank you!

$ ./init.sh
\n******************************************************************\n
Attempting to create anaconda environment....\n
Collecting package metadata (repodata.json): done
Solving environment: failed

ResolvePackageNotFound:

• scikit-image==0.13.0=np113py36_0
• icu==54.1=0
• sip==4.18=py36_0
• gst-plugins-base==1.8.0=0
• gstreamer==1.8.0=0
• mako==1.0.6=py36_0
• numpy==1.13.1=py36_0
• matplotlib==2.0.2=np113py36_0
• pytz==2017.2=py36_0
• freetype==2.5.5=2
• wheel==0.29.0=py36_0
• pip==9.0.1=py36_1
• markupsafe==1.0=py36_0
• appdirs==1.4.3=py36_0
• pywavelets==0.5.2=np113py36_0
• pillow==4.2.1=py36_0
• python==3.6.2=0
• libxml2==2.9.4=0
• hdf5==1.8.17=2
• libgcc==5.2.0=0
• libtiff==4.0.6=3
• setuptools==36.4.0=py36_1
• libgfortran==3.0.0=1
• pyqt==5.6.0=py36_2
• networkx==1.11=py36_0
• expat==2.1.0=0
• cudatoolkit==8.0=3
• scipy==0.19.1=np113py36_0
• jpeg==9b=0
• six==1.10.0=py36_0
• libxcb==1.12=1
• libffi==3.2.1=1
• numba==0.35.0=np113py36_0
• sqlite==3.13.0=0
• jbig==2.1=0
• readline==6.2=2
• certifi==2016.2.28=py36_0
• h5py==2.7.0=np113py36_0
• dbus==1.10.20=0
• mkl==2017.0.3=0
• openssl==1.0.2l=0
• decorator==4.1.2=py36_0
• tk==8.5.18=0
• python-dateutil==2.6.1=py36_0
• qt==5.6.2=5
• libiconv==1.14=0
• glib==2.50.2=1
• pcre==8.39=1
• fontconfig==2.12.1=3
• xz==5.2.3=0
• libpng==1.6.30=1
• olefile==0.44=py36_0
• zlib==1.2.11=0
• pyparsing==2.2.0=py36_0

Generating executable scripts...\n
3DFSC executable scripts have been generated for bash, csh, and sh.
Feel free to delete scripts that you will not use (e.g. csh and sh if you use bash).
You can now move the run3DFSC.sh executable script anywhere you like to run the program.
\n*****************************************************************\n

[joshua4289]

HI @jianhaoc this is simply saying that the packages listed above aren't available on conda. This is because the exact versions are quite old

[joshua4289]

In the same vein .
I have been trying to get 3DFSC to run with numba=0.34

Anisotropy/requirements.txt

Line 8 in 8000d20

numba==0.34.0

I am running on an RTX2080Ti on driver version 450.57 and CUDA version of 11.0

My steps are
1.I created a conda virtual env with python 3.6
2. pip install -r requirements.txt (from within the environment)

  File "/home/user/.conda/envs/3DFSC/lib/python3.6/site-packages/numba/cuda/cudadrv/nvvm.py", line 340, in __init__
    raise RuntimeError(MISSING_LIBDEVICE_FILE_MSG.format(arch=arch))
RuntimeError: Missing libdevice file for compute_50.
Please ensure you have package cudatoolkit 7.5.

3. I then download CUDA7.5 from nvidia and set NUMBAPRO_NVVM and NUMBAPRO_LIBDEVICE so numba can find a libdevice for compute_50

The issue is now that I get a

numba.cuda.cudadrv.error.NvvmError: Failed to compile

libnvvm : error: -arch=compute_75 is an unsupported option
NVVM_ERROR_INVALID_OPTION
libnvvm : error: -arch=compute_61 is an unsupported option
NVVM_ERROR_INVALID_OPTION

This as I understand it is because CUDA 7.5 does not know how to interface with the driver of my card . I have also tried enforce compatibility with NUMBA_FORCE_CUDA_CC ( I have checked from 2.0 to 7.5)

Is there a way to run 3DFSC with the newer GPU and recent cuda versions ? or possibly I have missed something