Tensorflow 1.15.5 can't sense GPU

Hi,

I am trying to install Tensorflow 1.15.5 . After the installation process they can’t sense the GPU, they only work in CPU mode. My system:
Jetson AGX Xavier
Ubuntu 18.04 .5 LTS
Jetpack 4.6

I try to install the two version of tensorflow from NVIDIA TensorFlow Container 21.07!
Enclosed you can see the installed components of Jetpack 4.6 cuda|690x388 (or at least these components were set at the installation)

What can be the problem? This is what I see in the python of the terminal:

from tensorflow.python.client import device_lib
2021-08-17 15:40:37.137139: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.2
WARNING:tensorflow:Deprecation warnings have been disabled. Set TF_ENABLE_DEPRECATION_WARNINGS=1 to re-enable them.

print(device_lib.list_local_devices())
2021-08-17 15:40:58.885020: W tensorflow/core/platform/profile_utils/cpu_utils.cc:98] Failed to find bogomips in /proc/cpuinfo; cannot determine CPU frequency
2021-08-17 15:40:58.886025: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x227d4c60 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2021-08-17 15:40:58.886115: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
2021-08-17 15:40:58.891663: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1
2021-08-17 15:40:58.969849: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1049] ARM64 does not support NUMA - returning NUMA node zero
2021-08-17 15:40:58.970255: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x21bb3070 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2021-08-17 15:40:58.970321: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Xavier, Compute Capability 7.2
2021-08-17 15:40:58.970727: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1049] ARM64 does not support NUMA - returning NUMA node zero
2021-08-17 15:40:58.970935: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1666] Found device 0 with properties:
name: Xavier major: 7 minor: 2 memoryClockRate(GHz): 1.377
pciBusID: 0000:00:00.0
2021-08-17 15:40:58.971006: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.2
2021-08-17 15:40:58.971273: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library ‘libcublas.so.10’; dlerror: libcublas.so.10: cannot open shared object file: No such file or directory
2021-08-17 15:40:58.971456: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library ‘libcufft.so.10’; dlerror: libcufft.so.10: cannot open shared object file: No such file or directory
2021-08-17 15:40:58.971621: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library ‘libcurand.so.10’; dlerror: libcurand.so.10: cannot open shared object file: No such file or directory
2021-08-17 15:40:58.971787: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library ‘libcusolver.so.10’; dlerror: libcusolver.so.10: cannot open shared object file: No such file or directory
2021-08-17 15:40:58.971989: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library ‘libcusparse.so.10’; dlerror: libcusparse.so.10: cannot open shared object file: No such file or directory
2021-08-17 15:40:58.972208: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library ‘libcudnn.so.8’; dlerror: libcudnn.so.8: cannot open shared object file: No such file or directory
2021-08-17 15:40:58.972243: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1689] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at जीपीयू समर्थन  |  TensorFlow for how to download and setup the required libraries for your platform.
Skipping registering GPU devices…
2021-08-17 15:40:58.972291: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1206] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-08-17 15:40:58.972316: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1212] 0
2021-08-17 15:40:58.972340: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1225] 0: N
[name: “/device:CPU:0”
device_type: “CPU”
memory_limit: 268435456
locality {
}
incarnation: 3278512751446374925
, name: “/device:XLA_CPU:0”
device_type: “XLA_CPU”
memory_limit: 17179869184
locality {
}
incarnation: 235154773731855864
physical_device_desc: “device: XLA_CPU device”
, name: “/device:XLA_GPU:0”
device_type: “XLA_GPU”
memory_limit: 17179869184
locality {
}
incarnation: 11167916173511087386
physical_device_desc: “device: XLA_GPU device”
]

KeyboardInterrupt

exit()

I appreciate any advice.

Hi,

Which container do you use?
For JetPack4.6 with TensorFlow 1.15.x, please use r32.6.1-tf1.15-py3 from the below NGC server:
https://ngc.nvidia.com/catalog/containers/nvidia:l4t-tensorflow/tags

Thanks

Thank you, the container that you linked is working fine!
The only point: is it somehow possible to connect a jupyter notebook to this container?

Hi,

Suppose yes.
You can install it via the following command:

$ pip3 install jupyter
$ jupyter notebook

Thanks.

Unfortunately pip3 or pip install doesn’t work for me on the Jetson AGX Xavier Device for jupyter notebook (at least not yet). I am very happy to hear any of your further suggestions to install it with other methods.

Jupyter notebook works already using:

$ pip3 install jupyter
$ jupyter notebook

Thank you for the help!