HEF version not supported, you will need to provide a config file

General Discussion
1

image
image1688×634 209 KB

i have train my dataset with yolov11n-seg, then i compile with yolov8n_seg by declaring the last end nodes, after having the hef file, i run the command on raspberry pi 5 and get the following problem, i don’t know how to fix it? can you help me?

image
image1416×101 5.77 KB

2

Hey @DevPhamPham,

Just checking—did you use the right JSON config file for the segmentation model? (These seg models need a JSON file for the config.)
Also, make sure you’re using the correct post-processing for your model.

To double-check that your .hef file is working, you can run:

hailortcli run {hef}

Let me know what you get!

1 Like
3

Hi @omria,

Thanks for replying, here is my result when running the command. And may I ask what JSON file should I provide?
and when compiling the yolov11n-seg model, I just need to do the same as yolov8n-seg but change the end-node names, right?

(hailodfc) devpham@DevPham:~$ hailortcli run rip_mask_yolov11n-seg_V1.3.hef
Running streaming inference (rip_mask_yolov11n-seg_V1.3.hef):
  Transform data: true
    Type:      auto
    Quantized: true
[HailoRT] [error] Can't find hailort driver class. Can happen if the driver is not installed, if the kernel was updated or on some driver failure (then read driver dmesg log)
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_DRIVER_NOT_INSTALLED(64) - Failed listing hailo devices
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_DRIVER_NOT_INSTALLED(64)
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_DRIVER_NOT_INSTALLED(64)
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_DRIVER_NOT_INSTALLED(64)
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_DRIVER_NOT_INSTALLED(64)
[HailoRT CLI] [error] CHECK_SUCCESS failed with status=HAILO_DRIVER_NOT_INSTALLED(64) - Failed creating vdevice
4

Hey,
This could be one of two things:

  1. Driver not installed
    To check that, run:

    hailortcli fw-control identify

    If this fails, the driver might not be installed or the device isn’t connected properly.

  2. Version compatibility issue
    Even if the driver is installed, mismatched versions between HailoRT, DFC, and HailoMZ can cause problems.
    can you share which versions you’re using for the following?

    • HailoRT
    • DFC (Dataflow Compiler)
    • Hailo Model Zoo (HailoMZ)
5

Hi, @omria ,

Can you help me? I’m having a similar issue. My model is based on yolov11-x-seg, and I don’t know anything about a JSON config file. Do I need to create it manually?

image
image951×475 63.9 KB

6

Hi @federico_Guerrero_Tr

Welcome to the Hailo community. At DeGirum (a SW partner of Hailo), we developed a cloud compiler to help users convert YOLO checkpoints to hef: Early Access to DeGirum Cloud Compiler. Please see if you find it helpful for your use case.

7

Hi @shashi

Thanks, I’ll definitely try it out while I wait for a reply to the original post.

8

Hi @shashi
I also have a similar problem, with a yolov11 segmentation network converted to HEF.

image
image1115×628 40.1 KB

I applied for an early access to DeGirum Cloud Compiler.
What I’m curious about is that DeGirum does mention support for semantic segmentation, but it doesn’t mention instance segmentation support?
Is there any additional information on config file available?"

9

Hi @Toni_Stojev

YOLO segmentation models which are instance segmentation models are supported by the cloud compiler. We will change the wording on the page to clarify this. Thanks for pointing it out.