CUDA Driver UNAVAILABLE (cuInit(0) returned 999)

AI & Data Science Deep Learning (Training & Inference) Frameworks
1

I have a Tesla P4 GPU driver 396.37 and i trying to use this image

FROM nvcr.io/nvidia/tensorflow:18.09-py3

But when run I get the following error

ERROR: This container was built for NVIDIA Driver Release 410 or later, but
version 396.37 was detected and compatibility mode is UNAVAILABLE.

   [[CUDA Driver UNAVAILABLE (cuInit(0) returned 999)]]

Is there any compatibility list, or can you help me a compatible image for my GPU

2

You have a few options, I’ll mention 2 of them.

1.The 18.08 container should work with your 396.37 driver. nvcr.io/nvidia/tensorflow:18.08-py3

  1. You can upgrade the driver on your machine to a 410.48 or newer driver. If you previously installed a driver using the runfile installer method, you should repeat that method with a newer driver. You can get that newer driver at [url]Official Drivers | NVIDIA If you previously installed the driver using a package manager method (e.g. sudo apt-get … ) then you should repeat that with a newer driver. In this case you could download CUDA 10 and install it. [url]http://www.nvidia.com/getcuda[/url] Use the deb install method and follow the instructions for deb install.

The driver versions for each container release are listed here:

[url]https://docs.nvidia.com/ngc/ngc-alibaba-vmi-release-notes/index.html[/url]