CUDA device not detected in Tensorflow

AI & Data Science Deep Learning (Training & Inference) TensorRT
, ,
1

I am having some issues with Tensorflow, that seems not to detect my GPU.

When running some code using Tensorflow, I get the error :

tensorflow/stream_executor/cuda/cuda_driver.cc:328]

failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected

Here’s my config :

  • Nvidia GeForce RTX 3080 Ti
  • Ubuntu 18.04
  • CUDA 11.4, driver 470.57.02
  • Tensorflow 2.5

My GPU is well detected (checked it with nvidia-smi ) and tf.test.is_gpu_available() returns True. I tried downgrading the CUDA version and the driver but nothing changed.

Does anybody has some hints on how to solve this? Thanks a lot!

2

Hi @g.san,

This issue looks like out of scope for TensorRT, we recommend you to post your concern on tensorflow related platform to get better help.
If you are interested, you can try tensorflow NGC container TensorFlow | NVIDIA NGC

Thank you.