yolo11n-pose optimisation error

@omria @Nadav @nina-vilela @pierrem hey, trying to convert yolo11n-pose model with [5, 2] keypoints and 4 classes, run command to optimize model and get:

File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_model_optimization/acceleras/model/hailo_model/model_flow.py”, line 181, in get_edge_output_index
return self.get_edge_data(u, v)[“output_index”]
TypeError: ‘NoneType’ object is not subscriptable

Before it, I had compiled another model with 6 classes and [4, 2] keypoints, and it was successful. Running with the use of the standard architecture of the model

Hey @ighgul!

This looks like it might be a configuration mismatch between your end nodes and keypoints. Since you’re working with a custom keypoint setup, the end nodes could be different from the default yaml configuration.

Quick question - are you using random calib data for testing?

Here are a couple of things that might help:

Let me know how it goes!

hey @omria , thank you for replying, I know about end nodes mismatch and I don’t see any suitable yaml files for 11n-pose, so I config them by myself, I can send you this info in dm

hey @omria , have some new details, i can say that error is raised from simple conv and concat. Info message is like ‘No shifts available for this conv layer’ and edge is between ne_activation_conv and this concat. I find some errors like this and change activation to sigmoid (just to debug) and I convert this model but as a result have

[HailoRT] [error] CHECK failed - Requested name for network_index=1, however there are only 1 networks in the network group
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_INVALID_ARGUMENT(2)

What is supposed to do?
@omria @Nadav @nina-vilela @pierrem