-
Notifications
You must be signed in to change notification settings - Fork 18.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improving Python packaging #4170
Comments
On the first point, I would like to clarify I also think Maybe if conda packages could be converted to wheels (though probably many changes on the wheels side of the fence) this might be possible, but that is a bit of a pipe dream. 💭 Sorry, I digress. Mainly I was think of using
Happy to see this is an idea you are willing to accept (at least according to the cited issue). Again any strategy that is picked here is an improvement. This could be done with |
I'm pretty sure the @BVLC team would welcome any pull requests you have to add a working |
As a follow-up on our previous conversation, @shelhamer, I have added this issue on packaging improvements. |
Any thoughts on this from the @BVLC team? I'd like to help improve this, but I don't want to start until we have some consensus about what needs to be done. |
For a windows conda package to be possible some dependencies are missing on conda-forge:
|
While that is an interesting question and I certainly would be interested to hear your thoughts, that is off topic w.r.t. this issue. The concern this issue is focused on is how to properly handle the installation of Python scripts. This is not an OS or conda-forge specific question. As you seem interested in the Windows packaging of Caffe on conda-forge, I invite you to share your thoughts in issue ( conda-forge/caffe-feedstock#3 ), @willyd . 😉 |
To echo @lukeyeager we would welcome pull requests for helping package up pycaffe the right way. |
I recently packaged
caffe
for conda-forge. I noticed a few things about the Python install that could be improved to make it easier for users to install and maintain the package thereof.One of the things I noticed is there wasn't a proper
setup.py
and that the Python files were not getting placed insite-packages
. Further the executables were not being placed in the right directory and lacked the right permissions.Another issue related to the executables were designed as if to be full-blown importable modules. In practice, it is better to separate out the executable portion into a small shim that gets installed elsewhere and has proper permissions. This can be done by either using
entry_points
or making your own shims. For doing the former, take a look at this example fromconda
itself. Here is an example of the latter with Jupyter'snotebook
.As this stuff can be a bit tedious to get right, I would recommend trying something like
cookiecutter
to get this all setup. This can bepip
installed. We also have itconda-forge
if you wish to go that route.The text was updated successfully, but these errors were encountered: