Re: [caffe] current status of packaging the deeplearning framework
On 07/08/15 09:55, lumin wrote:
On Fri, 2015-08-07 at 07:46 +0100, Ghislain Vaillant wrote:
I could not agree more. Please let us know if you need assistance in
puting the package to shape. I'd be happy to help.
Thank you :-)
Please consider the following changes:
[...]
Both done. Meta package is to be added.
It is not a requirement though. Just have a quick thinking whether this
meta-package is really needed or not.
However, I see no need to install caffe-cpu and caffe-cuda
together, as caffe-cuda can work at the cpu-only mode.
What are these packages supposed to contain ? Executables ? GUI apps ?
caffe-cuda needs to be linked with CUDA, while caffe-cpu needs
no CUDA. Hence I think pointing the meta package caffe/libcaffe
to caffe-cpu is proper, since Debian is universal, instead of
x86 / amd64 only. (CUDA is x86 / amd64 only)
I am not sure I am following you here.
I had a quick check and was wondering why you chose *not* to use the
CMake build system of caffe, unless I am mistaken?
[...]
The d/rules looks a bit convoluted and I would expect much simpler rules
if cmake were used.
The fact is, Makefile was fixed before CMake, and at that time
I'm only familiar with Make but not CMake. Besides, the patches
to Makefile contain some dirty hacks.
I agree to switch build system from Make to CMake, and will do it soon.
Feel free to checkout ArrayFire, clBLAS, clFFT, for examples of
CMake-using d/rules.
From the official instructions for caffe, it seems that you may require
an override for the tests, to run "make runtests" instead of the
standard "make tests". Otherwise, their CMake seems pretty standard and
all you need to figure out is the right CMake build options to pass to
dh_auto_configure.
Ghis
Reply to: