-
Notifications
You must be signed in to change notification settings - Fork 35
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
Not detected Intel Arc GPU even after installation following the manual #28
Comments
@teogi do you use TensorFlow 2.12? BTW, do you install the driver inside Ubuntu? https://github.com/intel/intel-extension-for-tensorflow/blob/main/docs/install/experimental/install_for_arc_gpu.md#native-linux-running-directly-on-hardware |
I do try on the native Ubuntu, but there is some problem with the version of Linux. |
btw, the downgrade to the TensorFlow 2.11 did not solve the problem unfortunately.
|
|
Now things getting more complicated on running the
Get a Segmentation fault without any verbose output. Seem like there is XPU available, but not well recognized. Another try on tensorflow with
No GPU devices were recognized. |
@teogi
|
seem like nothing is wrong with the log. some other information that I have seen in the tutorial might help is given below:
|
@teogi ITEX depends on sycl/opencl interfaces. |
Thank you for the documentation. I've been facing the same issue as OP. Both, under WSL2 and native Ubuntu 22.04. I hope the following info could help Graphic card Intel Arc a770 LE. Output of env_check.sh:
output of python -c "import intel_extension_for_tensorflow as itex; print(itex.version)" show the device:
Nevertheless when loading Tensorflow from a notebook shows "Can not found any devices":
clinfo output:
|
@teogi from you log, that only import ITEX can find GPU |
Sorry for the ignorance, how can I set I notice that the problem happens when I use jupyter notebook or jupyter lab. If I run a python script from the command line using the same environment it shows the GPU
If I don't enable the oneAPI componets through best regards |
@mikemayuare |
@NeoZhangJianyu Yes, you were right, I was using the default kernel. I set the kernel from the environment with itex (tf) with The problem still persist. I checked the json file to see if the kernel is pointing to the right path Best regards |
Thank you! |
Sorry for the late reply, here is my
Some information has provided since the beginning, to make it more clearly and thorough, I will requote it below with other information that asked for.
I have downgraded tensorflow from version 12 to 11 as advised; my ITEX version comes with 1.1.0, installed using |
btw, is there any way to turn off the Nvidia driver warnings when using ITEX?
I found it quite troublesome and misdirecting when come to identifying my true problem. Best Regards, |
@mikemayuare Please help to check the version of libstc++.so both in your conda env (eg path: If this does not help, could you please set the environment variables |
Sorry for the delay. @wangkl2 as you suspected, it was a library conflict issue. Thanks to all for the great support. |
@teogi Sorry repond delay. From you clinof output, the device is OK, so the GPU driver and OCL driver are OK.
While your example fail to get the device.
So it is likey some libray confiliction at APP/example level.
Thanks! |
We are sorry for this. Google Tensorflow output this as the default GPU is Nvidia GPU. Thanks! |
The conda libraries have the lib conflicts as you mentioned. The problem is the default version for anaconda is Solved the lib conflict problem by rebuild the environment through
also, there is some differences in the version of
The version at oneapi's env |
Hi @teogi , Sorry for the inconvenience.
|
If it helps, I installed |
@Ankur-singh, could you please try ZE_AFFINITY_MASK =1 python your_test.py ? This will mask the iGPU. I close the issues , but please feel free to reopen if any updates. FYI, there is new version for your future test: https://intel.github.io/intel-extension-for-tensorflow/latest/docs/install/experimental/install_for_arc_gpu.html |
Hi, I am facing an issue with the system identifying the Intel Arc GPU following the manual here: Experimental: Intel® Arc™ A-Series GPU Software Installation
Here is some spec of my system:
The issue that I have been facing since now is the last step
python -c "import intel_extension_for_tensorflow as itex; print(itex.__version__)"
which output
from the output, it seem that although I install the intel-extension for tensorflow , the output still requires CUDA driver.
By using the
intel_extension_for_tensorflow/tools/env_check.sh
in this github repo, I've passed all the tests provided, which means no problem detected over my system and dependency.here might be some other useful information provided by my system:
hwinfo
which looks like it didn't identify my GPU driver (it shows the vendor as Microsoft, instead of Intel as suggested? ).
I tried to find a solution to this but it didn't work out too well, all of the driver issues come with the cuda driver ):
The text was updated successfully, but these errors were encountered: