I want to use tensorflow with Jetson AGX Xavier and CUDA I installed OS and CUDA 10.2 using JETPACK 4.6.3
I tried below command and output was :
pip install tensorflow-gpu
Please install "tensorflow" instead.
So I did just 'pip install tensorflow' so tensorflow-cpu was installed.
I made the python file named 'check_gpu.py' for checking, which print GPU hardware
import tensorflow as tf
print( tf.config.list_physical_devices('GPU') )
check_device.py, which print all devices:
import tensorflow as tf
from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())
And I tried all below version;
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v46 tensorflow==2.6.2+nv21.12
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v46 tensorflow==1.15.5+nv21.12
I have three virtual-envs which have each python 3.8.0, 3.7.5, 3.6.9 and only 3.6 venv can installed tensorflow==2.6.2 & 1.15.5 because the others are not compatible.
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v46 tensorflow
This command installed latest version (tensorflow==2.11.0) But "check_device.py" didn't print any GPU. Only CPU was printed.
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v51 tensorflow==2.11.0+nv23.01
Output of "check_gpu.py" with tensorflow==2.11.0+nv23.1 was
Could not load dynamic library "libcudart.so.11.0": dlerror: ....
It can't find other library too. libcublas.so.11.0 , libcublasLt.so.11...etc.
I have only cuda-10.2, but it try to search cuda-11.0
How can I run the tensorflow-gpu function?
