-
Notifications
You must be signed in to change notification settings - Fork 1.5k
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 build branch r1.10 on windows #32
Comments
@guihigashi Do you execute those command in cmd.exe? I have tried bazel a long time ago and must be executed in bash (MSYS2 or git bash). Anyway, thank you very much, this protocol should be updated to the official documentation. |
I found it. They have updated the doc. |
I used cmd.exe. My guess is that bazel uses bash for some commands (git bash). |
Hello, I am looking forward to version 1.10 wheel. When will you release it? Thank you! |
I tried for two days to replicate guihigashi's method, which is nearly identical to the TensorFlow documentation, but it does not seem to work for me. The Bazel build completes fine, but it seems to get some things very wrong. For example, it compiles all the libraries as .so files even on Windows, when they should be dlls. Also, the bazel-bin\tensorflow\tools\pip_package\build_pip_package tmp/tensorflow_pkg command does not work because the simple_console_for_windows.zip created in bazel-bin\tensorflow\tools\pip_package after the build finishes is zero size, so the items required by the manifest (runfiles) are not being zipped up at all. Is anyone else able to replicate guihigashi's results and generate a working 1.10 wheel? |
I got a bunch of .so files too. I remember I tried newer bazel (0.16), that didn't work. The simple_console_for_windows.zip finished with 1.2GB. It was a bit of trial and error. I compile my wheel with CUDA capability 5.0, if you're interested I can share |
@guihigashi Can you share your 1.10?(windows gpu)I did not build successfully. |
@whb0219 https://github.com/guihigashi/tf-windows-wheel @fo40225 I apologize for sharing my repo here. If you want I can collaborate with this repo, but actually I plan to build future versions with the described config... Anyway, don't wanna be rude. |
@guihigashi Nevermind, Tensorflow is not my own library. |
@guihigashi Thank you for your help. |
I've been using this repo for some time and I wanted to share that I was able to build tf on windows 10 with bazel. This is based on the build log from the Official Builds.
Versions:
VS2017 with VS2015 workload
CUDA 9.2.148 and cuDNN 7.1.4
Python 3.6.6
Bazel 0.15.2
Clone the repo and checkout branch r1.10
VS prompt
Activate your python env
for CUPTI
for grep, bash and perl
bazel envs
this will create
.tf_configure.bazelrc
file with options you choose... cpu or gpu, cuda version, cudnn version,/arch:AVX2
or whatever...build command. (
--jobs 2 --local_resources 3000,2,1.0
is optional, I wanted to use the PC while building)Create wheel package
The text was updated successfully, but these errors were encountered: