-
Notifications
You must be signed in to change notification settings - Fork 74k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
bazel absence of ZIP64 support cause tensorflow build fail #22390
Comments
@meteorcloudy This is an interesting issue. |
The zip file is created by a custom zipper binary built from source in bazel, which doesn't have zip64 support. So installing zip64 from pacman is not going to help. The root cause here is we are zipping redundant files in I'm trying to figure out a way to exclude redundant files. |
Probably good to clean up what is going into that zip, but could also try to fix the zipper. Zlib seems to have zip64 support just fine, so maybe there is an issue with the bazel zlib client or something? I don't see any related issues on |
Yes, ideally bazel's zip tool should support zip64. I filed bazelbuild/bazel#6211 But I prefer to refactor tensorflow's dependency to clean up the zip file, because it will also improve the performance of creating TF pip package a lot. |
v1.11.0-rc2 + cherry-pick 2a01b6a It can build with most consumer compute capability. Thank you for help. |
@angersson if we are doing another RC, yes. |
Building the pip package succeeded for me after including the Error without "no dependencies" flag (Click to expand)(tf-gpu-src) C:\Users\hbeck3\tf-gpu\tensorflow>bazel-bin\tensorflow\tools\pip_package\build_pip_package C:\Users\hbeck3\tf-gpu\pip-package-cuda10-py3.6.6Sun Oct 28 13:50:53 WEST 2018 : === Preparing sources in dir: /tmp/tmp.cSTaMzHDFx (tf-gpu-src) C:\Users\hbeck3\tf-gpu\tensorflow> Success with "no dependencies" flag (Click to expand)(tf-gpu-src) C:\Users\hbeck3\tf-gpu\tensorflow>bazel build --define=no_tensorflow_py_deps=true --jobs 1 --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package[...] (tf-gpu-src) C:\Users\hbeck3\tf-gpu\tensorflow>bazel-bin\tensorflow\tools\pip_package\build_pip_package C:\Users\hbeck3\tf-gpu\pip-package-cuda10-py3.6.6 (tf-gpu-src) C:\Users\hbeck3\tf-gpu\tensorflow> |
@aaniin can you please post your setup as I'm still getting the same error even with Thanks! |
Hello @gunan, Would it be possible to add the flag recommendation here? https://www.tensorflow.org/install/source_windows I believe this would help many people who try to build Tensorflow from source for Windows. My goal was compute capability 3.5 support to have it run faster on my laptop. Here is the PR to docs repository: tensorflow/docs#178 Thank you! |
Thank you very much for your PR, it is reviewed, approved and merged! |
having same issue even after including build succeeds but following command gives error tried in both msys and cmd both showing same error
|
…enet/BUILD Address #22390#issuecomment-439610881 PiperOrigin-RevId: 223811309
I am getting this error during this both of these commands:-
I am not able to run build with bazel |
System information
Describe the problem
If build artifact large than 4GB, it can't generate the python package.
A easy way to reproduce the issue
related link:
#20332 (comment)
#22382
https://github.com/bazelbuild/bazel/blob/0.17.1/third_party/ijar/zip.cc#L74
Source code / logs
The text was updated successfully, but these errors were encountered: