Issue After Installing Hailo Dataflow Compiler

General Discussion
1

I’ve just completed the installation of the Hailo Dataflow Compiler, but I’m encountering an error.
I installed Dataflow Compiler by the following command in virtualenv.

pip3 install ./hailo_dataflow_compiler-3.33.0-py3-none-linux_x86_64.whl

I tried the following command, and recieved following message.

mycomputer@790:~$ hailo --help
[info] No GPU chosen, Selected GPU 1
Aborted (core dumped)

It seems that the GPU is not being recognized.
My environment is

  • Ubuntu 24.04
  • Nvidia Driver Version: 580.82.07
  • Cuda 12.8
  • Cuda Toolkit 12.5.82
  • CUDNN 9.13

Has anyone faced a similar issue or have suggestions on how to resolve this? Any help would be greatly appreciated!

2

I think There are some commands are to be used to get the result check “–help” on the terminal to get the correct command

3

@surya_yelamanchili
Thank you for your comment.
But, I encountered this error when I tried hailo --help command.

4

Hey @Mic_Ohara,

The error you’re encountering appears to be related to CUDA and CUDNN version incompatibilities.

The [info] No GPU chosen, Selected GPU 1 Aborted (core dumped) error indicates that the DFC is failing to properly initialize the GPU. Based on your system configuration, we’ve identified a version mismatch with the required dependencies.

Required Specifications for DFC v3.33.0:

  • CUDA Version: 12.5.1 (Required)
  • CUDNN Version: 9.10 (Required)
  • GPU Architecture: Pascal/Turing/Ampere/Ada

The newer versions of CUDA and CUDNN you’re running (12.8 and 9.13) have not been validated with DFC v3.33.0 and are likely causing the initialization failure.

Recommended Solution:

  1. Install exactly CUDA 12.5.1 and CUDNN 9.10 as specified in the official requirements
  2. Verify your environment variables (CUDA_HOME, LD_LIBRARY_PATH) point to the correct installation paths
  3. Confirm GPU visibility with nvidia-smi

Let me know if these help!