-
Notifications
You must be signed in to change notification settings - Fork 74.8k
Description
System information
- 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): Arch Linux
- TensorFlow installed from (source or binary): Source from branch r1.7, did not compile
- TensorFlow version (use command below): 1.7.0
- Python version: 3.6.4
- Bazel version (if compiling from source): 0.12.0
- GCC/Compiler version (if compiling from source): 7.3.1
- CUDA/cuDNN version: N/A
- GPU model and memory: N/A
- Exact command to reproduce: ./configure && bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package
Describe the problem
While compiling tensorflow build failed. I just pulled the latest commit(92e6c3e) from the r1.7 branch. I also updated bazel (from 0.11.1 to 0.12.0) and some other programs(should be irrelevant to tensorflow) beforehand.
Source code / logs
`ERROR: /home/UserHome/.cache/bazel/_bazel_UserHome/ab33c8274551e1ea3125872a4c4e7db9/external/jpeg/BUILD:126:12: Illegal ambiguous match on configurable attribute "deps" in @jpeg//:jpeg:
@jpeg//:k8
@jpeg//:armeabi-v7a
Multiple matches are not allowed unless one is unambiguously more specialized.
ERROR: Analysis of target '//tensorflow/tools/pip_package:build_pip_package' failed; build aborted:
/home/UserHome/.cache/bazel/_bazel_UserHome/ab33c8274551e1ea3125872a4c4e7db9/external/jpeg/BUILD:126:12: Illegal ambiguous match on configurable attribute "deps" in @jpeg//:jpeg:
@jpeg//:k8
@jpeg//:armeabi-v7a
Multiple matches are not allowed unless one is unambiguously more specialized.
INFO: Elapsed time: 1.086s
FAILED: Build did NOT complete successfully (3 packages loaded)`
UPDATE: The r1.8 branch compiles successfully. The 1.8 branch might have a fix? Or perhaps a dependency was updated and that fixed it?
Activity
Lu-Yi-Hsun commentedon Apr 12, 2018
try gcc 4~6 don't use gcc7
rharish101 commentedon Apr 12, 2018
I'm also facing build problems even with gcc-6.
System information
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): Arch Linux
TensorFlow installed from (source or binary): Source from master branch (r1.7)
TensorFlow version (use command below): 1.7.0
Python version: 3.6.4
Bazel version (if compiling from source): 0.11.1
GCC/Compiler version (if compiling from source): 6.4.1
CUDA/cuDNN version: 9.1.85/7.1.2
GPU model and memory: Nvidia GTX 960M, 4GB
Exact command to reproduce:
./configure && bazel build --config=opt --config=cuda --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" //tensorflow/tools/pip_package:build_pip_package --verbose_failures
Logs:
Moocow9m commentedon Apr 12, 2018
gcc is not the problem(tried 4-6). It has worked with gcc 7.3.1 before as well. The only changes to programs related to tensorflow is the tensorflow commit itself and bazel.
weinman commentedon Apr 12, 2018
To corroborate, I just experienced the same problem/error as in the original post. I'd compiled tf 1.7 successfully previously (Ubuntu 14.04.4) but just did an OS upgrade (Ubuntu 14.04.5) and now I get the same error. Several packages were updated---I'm not sure what may have been the cause.
dbkinghorn commentedon Apr 12, 2018
Getting back to Moocow9m's issue. I just had the same problem. It was the bazel update from a couple of days ago. Try un-installing version 12 and get the 11.1 bazel deb file and install form that. I also had to add --action_env PATH="$PATH" to keep it from forgetting where Python was (I'm using Anaconda3)
I also have the following cuda.conf in ld.so.conf.d
/usr/local/cuda/lib64
/usr/local/cuda/extras/CUPTI/lib64
/usr/local/cuda/targets/x86_64-linux/lib/stubs
Moocow9m commentedon Apr 12, 2018
@dbkinghorn It does compile with the older bazel build, however since version 0.12.0 can compile branch r1.8 I still think this might be a tensorflow issue in the r1.7 branch.
weinman commentedon Apr 13, 2018
I also can compile tf master with bazel 0.12. However, I'm still befuddled that the same tf commit (from branch 1.7) failed to compile after a suite of package updates. That doesn't necessarily mean it wasn't a tf issue in 1.7, just that it wasn't manifest until after a separate update (which I don't think changed bazel).
LuBingtan commentedon Apr 13, 2018
Maybe this problem is caused by bazel. I could compile branch r1.6 with bazel 0.11.1. If I update bazel from 0.11.1 to 0.12.0, it will be failed.
rharish101 commentedon Apr 13, 2018
I just pulled the latest commits, and it builds fine.
Moocow9m commentedon Apr 13, 2018
@rharish101 what version of bazel did you use? What is your OS?
ysfalo commentedon Apr 14, 2018
@LuBingtan , Yeah, must be the problem of bazel 0.12. I updated bazel from 0.12 to 0.11 and it seems fine now.
tensorflow 1.5
ubuntu16
cuda 8.0
cudnn 5.1
agupta74 commentedon Apr 14, 2018
Had similar issue building TF 1.7 with Bazel 0.12. Downgrading bazel to 0.11 seems to fix the issue
erickchandra commentedon Apr 14, 2018
Downgrading to Bazel 0.11.1 may fix this issue currently
-- Tensorflow r1.7 with CUDA 9.1 and CuDNN7.1
21 remaining items