Make Python/Numpy include paths configurable

The Mac OS X Numpy and Python include paths are hard-coded to /System/Library/Frameworks/Python.framework/Versions/2.7/Extras/lib/python/numpy/core/include and /usr/include/python2.7. I am on OS X 10.11 and I use the Anaconda Python distribution (installed to /Users/<user>/anaconda).

Currently, Bazel doesn't find my headers:

./tensorflow/python/client/tf_session_helper.h:4:10: fatal error: 'Python.h' file not found
#include <Python.h>
         ^
1 error generated.

It would be great if Python include paths could be configured in configure along with the cuDNN and CUDA paths.

I just ran into the same problem. A configuration would be great. Just for now I solved this by editing https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/BUILD#L23 and add your Python / Numpy include paths, e.g. for OS X "-I/System/Library/Frameworks/Python.framework/Versions/2.7/include/python2.7"

bf6b536 added the ability to configure the python paths -- as a consequence, we require users to now run ./configure before invoking bazel so they can specify the paths.

Let us know if this works for you!

[SYCL] Optimize fused batch norm (tensorflow#187)

Update Chapter 3 to demonstrate pattern match and rewrite optimizations

Merge pull request tensorflow#187 from parallelo/simplify-dockerfile

PR tensorflow#187: QConv3D Dequantize fusion

Navigation Menu

Repository navigation

Make Python/Numpy include paths configurable #187

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Make Python/Numpy include paths configurable #187

Description

Activity

tscheepers commented on Nov 17, 2015

vrv commented on Dec 3, 2015

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions