-
Notifications
You must be signed in to change notification settings - Fork 2k
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
Installation works for tfjs-node but failed for tfjs-node-gpu #1098
Comments
Is there a way to keep the .staging folder and @tensorflow/tfjs-node-gpu folder to help debugging out? Attached in is the log generated by the script, but I'm not sure how to dig deeper as all the scripts are cleaned up upon failure. I can post the --verbose log if needed, but it looks to have found all the correct paths and re-installing windows-build-tools didn't help. |
Another update: npm install --verbose @tensorflow/tfjs-node-gpu@0.2.1 works on my machine, but not 0.2.0, 0.2.2, or 0.2.3 |
Those logs have a truncated Do you mind cloning the tfjs-node repo and running the following - it might give some more details: git clone https://github.com/tensorflow/tfjs-node.git
cd tfjs-node
npm install
# This command probably fails:
npm run enable-gpu
# Get the logs from this:
node-gyp rebuild If you don't have node-gyp - Also one other question - which version of CUDA/CUDNN do you have - we have strict requirements that match TensorFlow 1.12: https://www.tensorflow.org/install/gpu |
PS C:\Users\mcelr\tfjs-node> node-gyp rebuild CUSTOMBUILD : error : ENOENT: no such file or directory, open 'C:\Users\mcelr\tfjs-node/deps/include/tensorflow/c/c_ |
You need to run |
Any luck building v1.0.1 with GPU support? I tried on
Also tried the same with Build Tools 2015 (v14), but I get the same error. All
|
Your system is not finding python2 - try the instructions here https://github.com/tensorflow/tfjs-node/blob/master/WINDOWS_TROUBLESHOOTING.md#missing-python2 |
I suggest setting NODE_OPTIONS environment to increase memory limit. I had success building tfjs-node v1.0.1 with GPU support on Windows 10/Build Tools 2015 using"NODE_OPTIONS=--max-old-space-size=4096" |
But it seems it has found it after searching in
I also made sure I can call python (python27) in cmd, is there anything else I need to do? |
Noticed that in my previous message the error was Is that the expected behaviour? |
So I tried I tried further with Build Tools 2015 (v14) and got
And
|
@saravanabalagi Sadly, the node-gyp build tool uses gyp which has had various issues (just search issues on that repo) keeping up with Visual Studio build tool updates. Chrome moved away from gyp to gn a couple of years ago. While there are efforts underway to help, it's probably best for us to start providing users with pre-built binaries (at least on Windows to start). Please follow this issue: #1390 |
@nkreeger, FWIW, I finally got tfjs-node-gpu to install on Win10. I had to install version 0.3.2 for it to work. Trying any of the 1.0.x versions would result in the above mentioned node-gyp rebuild failures. Any possibility that this is a similar issue to the gpu packaging / layout issue with the upstream version of TF? |
With the latest tfjs-node/tfjs-node-gpu package, we are shipping pre-compile addon and there should be no compilation on your machine. This issue should have been fixed. |
I got this error when installing
npm i @tensorflow/tfjs-node-gpu
that some people have also gotten. I've tried the solutions that worked for most people, but still nothing worked for me.I tried
npm i -g --production windows-build-tools
, also locally. I tried installing windows-build-tools manually as followed from the node-gyp github page.I tried to install the CPU (
npm i @tensorflow/tfjs-node
) version and it was installed with no problems.I have CUDA, CuDNN, etc installed as instructed.
I would really appreciate if someone could help me out with this.
Windows 10
Python 2.7.15
Node v10.15.0
npm v6.4.1
node-gyp 3.8.0
The text was updated successfully, but these errors were encountered: