2 issues (isinstance(), ConvolutionLayer()) when installing barrista about barrista HOT 4 CLOSED

First, I found something strange part like this:
('CAFFE_PYTHON_FOLDER', '/home/young/.conda/pkgs/pytest-3.6.2-py27_0/lib/python2.7/site-packages/caffe')
I think pytest-3.6.2-py27_0 part should be py27caffegpu which is name of conda environment.
I'm going to check it first.

from barrista.

I fixed the paths but issues look same

from barrista.

Hi youngmtool!

Unfortunately, barrista is unmaintained with the rise of the pythonic widely spread deep learning frameworks. If you still want/need to use it, as I figure is your case, I recommend using precisely the caffe version linked (however, this may make problems with more recent versions of CUDA :( ), if necessary within a docker image. There should still be Docker images floating around with CUDA7 and corresponding CuDNN versions.

I'm not sure what your first reported error comes from. Commenting out those lines is not an option, this leads to the other failures (barrista will not have parsed the layer properties). If you don't already, please use Python 2.7.

from barrista.

Ah, right. Thanks for reply and all information. It helped a lot. I'll try to follow your advice.

from barrista.