提交 d4d9be53 authored 作者: James Bergstra's avatar James Bergstra

modified install.txt to reflect new config options for blas

上级 2e323b39
......@@ -86,48 +86,31 @@ In csh:
setenv PYTHONPATH <path to Theano's parent dir>/Theano:$PYTHONPATH
Configuring Theano's environmental variables
---------------------------------------------
Two environment variables are used to control automatic code
generation. It is possible to use Theano in a way which avoids all
automatic code generation, but that way is much, much slower.
- ``THEANO_BLAS_LDFLAGS``:
A space-separated list of options (flags) that will be passed to the
compiler. It should include library names to link against for BLAS
functions. Default: ``-lblas``.
- ``THEANO_BASE_COMPILEDIR``:
A directory with read/write access permissions where Theano will store
a sub directory for each architecture for autogenerated code and C modules.
Default: ``$HOME/.theano/``. The sub directory is name as
``compiledir_xxx`` where ``xxx`` is a platform identifier.
If this
directory does not exist or does not have the correct permissions,
Theano will try to create it with the correct permissions. If that fails,
an exception will be raised and no C code will be compiled.
- ``THEANO_COMPILEDIR``:
A directory with read/write access permissions where Theano will store
autogenerated code and C modules. Default:
``$HOME/.theano/compiledir_xxx`` where ``xxx`` is a platform identifier.
If this
directory does not exist or does not have the correct permissions,
Theano will try to create it with the correct permissions. If that fails,
an exception will be raised and no C code will be compiled.
- ``THEANO_DEFAULT_MODE``:
String value specifying the default mode to use when compiling Theano
graphs. This can be one of the strings defined in
:ref:`using_modes`.
Possible values so far are:
- ``'FAST_COMPILE'``
- ``'FAST_RUN'``
- ``'DEBUG_MODE'``
Omitting this variable defaults the mode to ``'FAST_RUN'``.
Troubleshooting: Make sure you have a BLAS library
-----------------------------------------------------
Theano currently depends on having a BLAS library to link against. The
default is to use the blas installation information in numpy:
``numpy.distutils.__config__.show``, but if numpy is not configured
[correctly] for using a fast BLAS, you should tell Theano which BLAS to use
via the THEANO_FLAGS environment variable, or a ``.theanorc`` file.
The .theanorc file is the simplest way to set a relatively permanent option
like this one. Add a ``[blas]`` section with an ``ldflags``
entry like this:
.. code-block:: text
# other stuff can go here
[blas]
ldflags = -lf77blas -latlas #put your flags here
# other stuff can go here
If you prefer to use an environment variable, type something like this: ``export THEANO_FLAGS="blas.ldflags=<gcc linker flags>"``
so maybe ``export THEANO_FLAGS="blas.ldflags=-lf77blas -latlas"`` might work for you.
Testing your installation
---------------------------
......@@ -217,9 +200,6 @@ Mac
- ``scipy`` depends on ATLAS (a BLAS library), which will be installed by MacPorts.
- Finally, set ``THEANO_BLAS_LDFLAGS`` to something which will link against said BLAS
library. E.g., ``THEANO_BLAS_LDFLAGS='-lcblas -latlas -lgfortran'``.
These installation instructions have not tested recently, please infom us of your results!
We would be especially interested in dependencies that we missed listing, as well as tests
that fail on your platform (use the ``theano-users@googlegroups.com`` mailing list).
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论