VStream related error

General
,
1

Hey Everyone,
I’m new to the community and new to using hailo. I had trained my custom model using this guide

However when I try to run it I am getting a
“VStream best/conv52 not found in sorted output names” and
“CHECK_SUCCESS failed with status=HAILO_INTERNAL_FAILURE(8)” error

Not sure what it means, nor there are any other guides on that. I appreciate any help coming towards me

Thanks & Regards

Hardware : Raspberry pi 5 4gb, Hailo 8L
Software : Debian 12 Bullseye, HailoRT & Other hailo versions @ 4.20

2

Hey @Nitish_K ,

Welcome to the Hailo Community!

What the errors mean

  • “VStream best/conv52 not found in sorted output names”
    Your pipeline is looking for an output node called best/conv52, but your model doesn’t actually have one by that name. This usually means your post-processing code or config expects a different output node than what your model is providing.
  • “CHECK_SUCCESS failed with status=HAILO_INTERNAL_FAILURE(8)”
    This is a generic runtime error, usually just telling you something already went wrong (like above).

Why it happens

  1. Custom YOLO Models:
    If you retrained YOLO (like YOLOv11n) and changed the number of classes, the output node name might not match what is in the example.
  2. Model Compilation:
    If you didn’t specify the right output nodes during compilation, or used the wrong config, the output names get mismatched.

How to fix it

1. Find out your actual output node names

import onnx
m = onnx.load('your_model.onnx')
print([o.name for o in m.graph.output])
  • See what the output node(s) are actually called.
  • Make sure your code or pipeline is asking for these, not something like best/conv52 unless your model really has that.

2. Update your post-processing/inference code

  • Change the output node names in your code/config to match what your model actually uses.
  • For Hailo’s Model Zoo or example apps, they sometimes expect a specific name—make sure it matches.

3. Recompile if you need to

  • If you need a different output node name, you can recompile your model for Hailo and specify the output nodes:
hailomake --onnx your_model.onnx --output-nodes <correct_output_name>