I’m having trouble converting an ONNX model to HEF format.
Here’s my issue:
. Installed Hailo RT PCIE driver 4.19.0 (Ubuntu) from the official website.
. Installed Hailo8_FW.4.19.0.bin.
. Built and followed Hailo documentation.
. hailo-compiler command not found.
. Running Ubuntu 22.4 with Python 3.10.
. Docker attempt failed due to same PCIE driver requirement.
Could you please assist with:
. Alternative PCIE driver versions compatible with Ubuntu 22.4?
. Alternative methods to convert ONNX to HEF?
It sounds like there might be some confusion with the command for the Dataflow Compiler (DFC). The correct command is hailo, not hailo-compiler. However, to simplify the setup, we highly recommend using our AI SW Suite Docker. This Docker environment has everything pre-installed, including the compiler, so you can easily run the conversion from within the Docker.
Alternatively, if you prefer, you can use the Hailo Model Zoo, which also leverages the compiler and provides a more user-friendly interface for working with models. Either option should help streamline the ONNX to HEF conversion process for you!
Even when trying with Hailo, I’m still getting a “command not found” error. I’ll try using the AI Software Suite Docker instead, but my system has NVIDIA-SMI 535.183.01, CUDA 12.2, cuDNN 9.1.0, Ubuntu 22.04, and Python 3.10. Could this be causing the issue? If not, what should I keep in mind when working with Docker?