Hi,
I am trying to use MXNet in Spaces and I have several questions on MXNet support.
I put mxnet-cu102mkl==1.6.0.post0
in the requirements.txt
and prepared app.py
which (indirectly) imports MXNet.
When pushed my app to Spaces, I got a runtime error, as follows:
Traceback (most recent call last):
File "app.py", line 12, in <module>
... (omitted)
File "/home/user/app/coordinate_reg/image_infer.py", line 4, in <module>
import mxnet as mx
File "/home/user/.local/lib/python3.8/site-packages/mxnet/__init__.py", line 24, in <module>
from .context import Context, current_context, cpu, gpu, cpu_pinned
File "/home/user/.local/lib/python3.8/site-packages/mxnet/context.py", line 24, in <module>
from .base import classproperty, with_metaclass, _MXClassPropertyMetaClass
File "/home/user/.local/lib/python3.8/site-packages/mxnet/base.py", line 214, in <module>
_LIB = _load_lib()
File "/home/user/.local/lib/python3.8/site-packages/mxnet/base.py", line 205, in _load_lib
lib = ctypes.CDLL(lib_path[0], ctypes.RTLD_LOCAL)
File "/usr/local/lib/python3.8/ctypes/__init__.py", line 373, in __init__
self._handle = _dlopen(self._name, mode)
OSError: libcublas.so.10: cannot open shared object file: No such file or directory
I assume there are no CUDA related files below /usr/local/
or any other paths.
My question are:
(1) Is it possible to somehow use MXNet in Spaces (without causing errors)?
(2) If not, how do I build a customized container environment to suit my app’s requirements?
I once saw a description that says I should contact the team to do so, is it the right way?