Syntaxnet fails to compile with XLA enabled on MacOS

[udnaan]

System information

• What is the top-level directory of the model you are using:
  models/syntaxnet
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
  No
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
  MacOS 10.12.4
• TensorFlow installed from (source or binary):
  Source
• TensorFlow version (use command below):
  a7d6015
• Bazel version (if compiling from source):
  0.4.5 (brew)
• CUDA/cuDNN version:
  8.0/5.1
• GPU model and memory:
  1080 TI 11GB
• Exact command to reproduce:

$ ./configure
cd ..
 bazel test --linkopt=-headerpad_max_install_names \
    dragnn/... syntaxnet/... util/utf8/...

Describe the problem

See subject

Source code / logs

Please specify the location of python. [Default is /usr/local/bin/python]:
Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]:
Do you wish to build TensorFlow with Google Cloud Platform support? [y/N]
No Google Cloud Platform support will be enabled for TensorFlow
Do you wish to build TensorFlow with Hadoop File System support? [y/N]
No Hadoop File System support will be enabled for TensorFlow
Do you wish to build TensorFlow with the XLA just-in-time compiler (experimental)? [y/N] y
XLA JIT support will be enabled for TensorFlow
Found possible Python library paths:
  /usr/local/Cellar/python/2.7.13/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages
Please input the desired Python library path to use.  Default is [/usr/local/Cellar/python/2.7.13/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages]

Using python library path: /usr/local/Cellar/python/2.7.13/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages
Do you wish to build TensorFlow with OpenCL support? [y/N]
No OpenCL support will be enabled for TensorFlow
Do you wish to build TensorFlow with CUDA support? [y/N] y
CUDA support will be enabled for TensorFlow
Please specify which gcc should be used by nvcc as the host compiler. [Default is /usr/bin/gcc]:
Please specify the CUDA SDK version you want to use, e.g. 7.0. [Leave empty to use system default]: 8.0
Please specify the location where CUDA 8.0 toolkit is installed. Refer to README.md for more details. [Default is /usr/local/cuda]:
Please specify the Cudnn version you want to use. [Leave empty to use system default]: 5
Please specify the location where cuDNN 5 library is installed. Refer to README.md for more details. [Default is /usr/local/cuda]:
Please specify a list of comma-separated Cuda compute capabilities you want to build with.
You can find the compute capability of your device at: https://developer.nvidia.com/cuda-gpus.
Please note that each additional compute capability significantly increases your build time and binary size.
[Default is: "3.5,5.2"]: 6.1
INFO: Starting clean (this may take a while). Consider using --expunge_async if the clean takes more than several minutes.
........
INFO: All external dependencies fetched successfully.
Configuration finished
$ cd ..
$ bazel test --linkopt=-headerpad_max_install_names \
                                             dragnn/... syntaxnet/... util/utf8/...
WARNING: /private/var/tmp/_bazel_nan/a58a8744e0a370eb1d612194d68aa421/external/org_tensorflow/tensorflow/workspace.bzl:72:5: tf_repo_name was specified to tf_workspace but is no longer used and will be removed in the future.
ERROR: /private/var/tmp/_bazel_nan/a58a8744e0a370eb1d612194d68aa421/external/llvm/BUILD:176:1: no such package 'third_party/llvm': BUILD file not found on package path and referenced by '@llvm//:llvm_config_gen'.
ERROR: Analysis of target '//dragnn/python:biaffine_units' failed; build aborted.
INFO: Elapsed time: 73.200s
ERROR: Couldn't start the build. Unable to run tests.

[drpngx]

Could you run bazel fetch ... again? Which version of tensorflow is that? We removed bazel clean during configure a while back.

[udnaan]

The commit tag is in the report. It's the version that works with syntaxnet and is provided as git submodule.

I'm following the instructions here to run syntaxnet: https://github.com/tensorflow/models/tree/master/syntaxnet#installation

I did try running it using master/HEAD beforehand but that seems to be broken altogether.

[udnaan]

Here is the error I get if I try using tensorflow from master branch with syntaxnet (note that syntaxnet itself is still using the git-tag mentioned in the report as it fails to compile with the master)

---------------------------------------------------------------------------
NotFoundError                             Traceback (most recent call last)
<ipython-input-1-9239b98b75e1> in <module>()
      6 from dragnn.python import graph_builder
      7 from dragnn.python import spec_builder
----> 8 from dragnn.python import load_dragnn_cc_impl  # This loads the actual op definitions
      9 from dragnn.python import render_parse_tree_graphviz
     10 from dragnn.python import visualization

/usr/local/lib/python2.7/site-packages/dragnn/python/load_dragnn_cc_impl.py in <module>()
     20 
     21 tf.load_op_library(
---> 22     os.path.join(tf.resource_loader.get_data_files_path(), 'dragnn_cc_impl.so'))

/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/load_library.pyc in load_op_library(library_filename)
     62       # pylint: disable=protected-access
     63       raise errors_impl._make_specific_exception(
---> 64           None, None, error_msg, error_code)
     65       # pylint: enable=protected-access
     66   finally:

NotFoundError: dlopen(/usr/local/lib/python2.7/site-packages/dragnn/python/dragnn_cc_impl.so, 6): Symbol not found: __ZN10tensorflow15shape_inference16InferenceContext8WithRankENS0_11ShapeHandleEiPS2_
  Referenced from: /usr/local/lib/python2.7/site-packages/dragnn/python/dragnn_cc_impl.so
  Expected in: flat namespace
 in /usr/local/lib/python2.7/site-packages/dragnn/python/dragnn_cc_impl.so

demangled name of the Symbol: _tensorflow::shape_inference::InferenceContext::WithRank(tensorflow::shape_inference::ShapeHandle, int, tensorflow::shape_inference::ShapeHandle*)

[drpngx]

Can you rebuild dragnn against TF master?

Here's the signature I have from the current codebase:

  Status WithRank(ShapeHandle shape, int64 rank,
                  ShapeHandle* out) TF_MUST_USE_RESULT;

It's int64, not int32.

[udnaan]

Giving it a go. Note that the build only goes through if I omit the MacOS specific option --linkopt=-headerpad_max_install_names specified in the Syntaxnet build instructions. Seems like it's not a valid option anymore.

[udnaan]

That seems to work. It also takes care of #248 if you separately built and install tensorflow with CUDA and XLA support
Thank you for the help!

[drpngx]

Yay! Thanks for following up.

…

On May 6, 2017 9:31 PM, "Adnan" ***@***.***> wrote: Closed #1444 <#1444>. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1444 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AT_Sbe-5VyWAKaRgyV6m4b1YE2ISYS7wks5r3Uk6gaJpZM4NRkYE> .