%global pkgname Theano
%global srcname theano
#%%global rctag a1
Name: python-theano
Version: 0.8.0
Release: 1%{?rctag:.%{rctag}}%{?dist}
Summary: Mathematical expressions involving multidimensional arrays
License: BSD
URL: http://deeplearning.net/software/theano/
Source0: https://github.com/Theano/Theano/archive/rel-%{version}%{?rctag:%{rctag}}/%{pkgname}-%{version}%{?rctag:%{rctag}}.tar.gz
# Fix some documentation bugs
Patch0: %{name}-doc.patch
BuildArch: noarch
BuildRequires: atlas-devel
BuildRequires: epydoc
BuildRequires: gcc-c++
BuildRequires: gcc-gfortran
BuildRequires: pydot
BuildRequires: python2-Cython python3-Cython
BuildRequires: python2-devel python3-devel
BuildRequires: python2-nose python3-nose
BuildRequires: python2-numpy python3-numpy
BuildRequires: python2-scipy python3-scipy
BuildRequires: python2-setuptools python3-setuptools
BuildRequires: python2-six python3-six
BuildRequires: python2-sphinx
BuildRequires: tex(latex)
BuildRequires: tex-dvipng
%description
Theano is a Python library that allows you to define, optimize, and
evaluate mathematical expressions involving multi-dimensional arrays
efficiently. Theano features:
- tight integration with NumPy: Use numpy.ndarray in Theano-compiled
functions.
- transparent use of a GPU: Perform data-intensive calculations up to
140x faster than with CPU.(float32 only)
- efficient symbolic differentiation: Theano does your derivatives for
function with one or many inputs.
- speed and stability optimizations: Get the right answer for log(1+x)
even when x is really tiny.
- dynamic C code generation: Evaluate expressions faster.
- extensive unit-testing and self-verification: Detect and diagnose many
types of mistake.
%package -n python2-%{srcname}
Summary: Mathematical expressions involving multidimensional arrays
Requires: atlas-devel
Requires: gcc-c++
Requires: gcc-gfortran
Requires: pydot
Requires: python2-numpy
Requires: python2-six
Requires: python2-scipy
%{?python_provide:%python_provide python2-%{srcname}}
%description -n python2-%{srcname}
Theano is a Python library that allows you to define, optimize, and
evaluate mathematical expressions involving multi-dimensional arrays
efficiently. Theano features:
- tight integration with NumPy: Use numpy.ndarray in Theano-compiled
functions.
- transparent use of a GPU: Perform data-intensive calculations up to
140x faster than with CPU.(float32 only)
- efficient symbolic differentiation: Theano does your derivatives for
function with one or many inputs.
- speed and stability optimizations: Get the right answer for log(1+x)
even when x is really tiny.
- dynamic C code generation: Evaluate expressions faster.
- extensive unit-testing and self-verification: Detect and diagnose many
types of mistake.
%package -n python3-%{srcname}
Summary: Mathematical expressions involving multidimensional arrays
Requires: atlas-devel
Requires: gcc-c++
Requires: gcc-gfortran
Requires: python3-numpy
Requires: python3-scipy
Requires: python3-six
%{?python_provide:%python_provide python3-%{srcname}}
%description -n python3-%{srcname}
Theano is a Python library that allows you to define, optimize, and
evaluate mathematical expressions involving multi-dimensional arrays
efficiently. Theano features:
- tight integration with NumPy: Use numpy.ndarray in Theano-compiled
functions.
- transparent use of a GPU: Perform data-intensive calculations up to
140x faster than with CPU.(float32 only)
- efficient symbolic differentiation: Theano does your derivatives for
function with one or many inputs.
- speed and stability optimizations: Get the right answer for log(1+x)
even when x is really tiny.
- dynamic C code generation: Evaluate expressions faster.
- extensive unit-testing and self-verification: Detect and diagnose many
types of mistake.
%package doc
Summary: Theano documentation
%description doc
User documentation for Theano.
%prep
%setup -q -n %{pkgname}-rel-%{version}%{?rctag:%{rctag}}
%patch0
# Remove bundled python-six
rm -f theano/compat/six.py
# Prepare for python 3 build
cp -a . %{py3dir}
mv %{py3dir} python3
# We don't need to use /usr/bin/env
for fil in $(grep -FRl /usr/bin/env python3); do
sed 's,/usr/bin/env[[:blank:]]*python.*,/usr/bin/python3,' $fil > $fil.new
touch -r $fil $fil.new
chmod a+x $fil.new
mv -f $fil.new $fil
done
for fil in $(grep -FRl /usr/bin/env .); do
sed 's,/usr/bin/env[[:blank:]]*python.*,/usr/bin/python2,' $fil > $fil.new
touch -r $fil $fil.new
chmod a+x $fil.new
mv -f $fil.new $fil
done
%build
# Regenerate the Cython files, and fix the numpy interfaces
cython theano/scan_module/scan_perform.pyx
pushd python3
cython3 theano/scan_module/scan_perform.pyx
popd
sed -e 's/\(__pyx_v_self\)->descr/PyArray_DESCR(\1)/' \
-e 's/\(__pyx_v_arr\)->base = \(.*\);/PyArray_SetBaseObject(\1, \2);/' \
-e 's/\(__pyx_v_arr\)->base/PyArray_BASE(\1)/' \
-i theano/scan_module/scan_perform.c \
python3/theano/scan_module/scan_perform.c
# Python 2 build
%py2_build
# Python 3 build
pushd python3
%py3_build
popd
# Build the documentation
export PYTHONPATH=$PWD
python2 doc/scripts/docgen.py --nopdf
# Remove build artifacts
rm -fr html/.buildinfo html/.doctrees
%install
# Install python 2 build
%py2_install
# Install python 3 build
pushd python3
%py3_install
popd
# Restore executable permission on the scripts
chmod a+x $(find %{buildroot} -name \*.py -o -name \*.sh | xargs grep -l '^#!')
%check
PYTHONPATH=$PWD bin/theano-test
%files -n python2-%{srcname}
%doc DESCRIPTION.txt HISTORY.txt NEWS.txt README.txt
%license doc/LICENSE.txt
%{python2_sitelib}/*
%files -n python3-%{srcname}
%doc DESCRIPTION.txt HISTORY.txt NEWS.txt README.txt
%license doc/LICENSE.txt
%{_bindir}/theano-*
%{python3_sitelib}/*
%files doc
%doc html
%changelog
* Wed Mar 23 2016 Jerry James <loganjerry@gmail.com> - 0.8.0-1
- New upstream release
* Thu Feb 04 2016 Fedora Release Engineering <releng@fedoraproject.org> - 0.7.1-0.2.a1
- Rebuilt for https://fedoraproject.org/wiki/Fedora_24_Mass_Rebuild
* Mon Feb 1 2016 Jerry James <loganjerry@gmail.com> - 0.7.1-0.1.a1
- Comply with latest python packaging guidelines
* Thu Nov 12 2015 Igor Gnatenko <i.gnatenko.brain@gmail.com> - 0.7.1-0.1.a1
- Update to 0.7.1a1
* Tue Nov 10 2015 Fedora Release Engineering <rel-eng@lists.fedoraproject.org> - 0.7.0-2.2
- Rebuilt for https://fedoraproject.org/wiki/Changes/python3.5
* Wed Nov 4 2015 Toshio Kuratomi <toshio@fedoraproject.org> - 0.7.0-2.1
- Fix python3 package requiring python2.
* Thu Jun 18 2015 Fedora Release Engineering <rel-eng@lists.fedoraproject.org> - 0.7.0-1.1
- Rebuilt for https://fedoraproject.org/wiki/Fedora_23_Mass_Rebuild
* Wed Apr 1 2015 Jerry James <loganjerry@gmail.com> - 0.7.0-1
- New upstream release
- Drop upstreamed -arm patch
- Regenerate cython files to fix build failure
* Sat Feb 21 2015 Jerry James <loganjerry@gmail.com> - 0.6.0-5
- Add -arm patch to fix build failure on arm builders due to inverted test
* Sat Feb 21 2015 Jerry James <loganjerry@gmail.com> - 0.6.0-4
- Drop workaround for fixed bug (bz 1075826)
- Use license macro
* Sat Jun 07 2014 Fedora Release Engineering <rel-eng@lists.fedoraproject.org> - 0.6.0-3.1
- Rebuilt for https://fedoraproject.org/wiki/Fedora_21_Mass_Rebuild
* Wed May 14 2014 Jerry James <loganjerry@gmail.com> - 0.6.0-3
- Rebuild for https://fedoraproject.org/wiki/Changes/Python_3.4
* Thu Mar 13 2014 Jerry James <loganjerry@gmail.com> - 0.6.0-2
- Add python3 subpackage
- Add another icon to the -missing tarball
- Update source icons
- Unbundle python-six
- Add workaround for bz 1075826
* Sat Dec 7 2013 Jerry James <loganjerry@gmail.com> - 0.6.0-1
- New upstream release
- Drop upstreamed -import patch
* Mon Oct 21 2013 Jerry James <loganjerry@gmail.com> - 0.6.0-0.1.rc3
- Add the -import patch to fix an exception
- Add more files to the base package docs
* Tue Aug 27 2013 Jerry James <loganjerry@gmail.com> - 0.6.0-0.rc3
- Initial RPM