"Could not find cuda drivers on your machine, GPU will not be used" with tensorflow

TensorFlow General Discussion
1

0

Hi I am trying to use my GPU and tensorflow, but it does not let me:

Error message: “Could not find CUDA drivers on your machine”

image
image723×332 130 KB

OS: Ubuntu 20.04, Conda environment
Graphic card: NVIDIA GeForce GTX 1660 Ti

image
image816×509 106 KB

Driver version: 520.61.05

image
image804×457 141 KB

Cuda version: 11.8

image
image529×139 53 KB

Tensorflow version: 2.13.1

image
image814×245 134 KB

Conda version: 23.9.0
Python version: 3.11.5

Thanks

4

I cannot reproduce the issue here, but if you are OK to upgrade Tensroflow to v2.14, please try out: pip3.11 install tensorflow[and-cuda]==2.14

It looks like TF2.14 comes up with the list of required appropriate cuda, cudnn and Tensorrt packages to install along Tensorflow. See here. I appears this was not in v2.13.

6

thanks seems to work but still throwing a lot of warnings, looks like Tensorflow has not been fully implemented for this type of Graphic Card:

2023-10-28 00:38:09.383818: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:9342] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered 2023-10-28 00:38:09.383871: E tensorflow/compiler/xla/stream_executor/cuda/cuda_fft.cc:609] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered 2023-10-28 00:38:09.384755: E tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:1518] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered 2023-10-28 00:38:09.469252: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-10-28 00:38:10.814586: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

Using GPU

2023-10-28 00:38:12.236693: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:894] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at nux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 2023-10-28 00:38:12.446312: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:894] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at http
tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:894] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at ht