Your Chosen Target Does Not Support this model version

,
#1

Topic:
Trying to BYOM into edge impulse to make comilation of provided onnx files easier

Project ID:
[Provide the project ID, if applicable]

Context/Use case:
Using a basic, small, onnx model, i want to use edge impulse to test how well it transfers my onnx file to my hardware (TI AM62A). Since this hardware is part of the supported devices family, I thought it would be a good test.

Details:
I uploaded the onnx file to the BYOM pipeline, including a supporting npy calibration file, but got the following error:
Your chosen target (TI AM62A (with Deep Learning Accelerator)) does not support this model version.

I have previously deployed this onnx file using TI’s onnxruntime-tidl tools before and had some basic success, so have a baseline that it is a model that can run on this device

Questions:
Not sure what the error means or how to get more details other than the model is “not supported”

It doesn’t look like i can upload the onnx file here, otherwise i would.

Thanks!

#2

Hi @lrid-val - welcome to the forum!

Could you please clarify at what point in the workflow that you are seeing the error message: Your chosen target (TI AM62A (with Deep Learning Accelerator)) does not support this model version.? Also perhaps what model you are using?

I grabbed a random computer vision model from the examples in the ONNX repo (onnx/models/Computer_Vision) and uploaded it using BYOM with the target selected as TI AM62A. I did not upload representative features for quantization. I was able to successfully upload the model and was unable to reproduce the error.

After uploading the model, I did run into some other errors on the Deployment page when building the TIDL-RT library and the .eim file for the TI AM62A, based on some operator support. Different issue from what you are reporting though.

#3

Thank you Brian!

Step 1: Upload pretrained model - Step 1: Upload a model.

  • In the first step of the BYOM workflow, i upload the onnx file.
  • I click “Upload file” and proceed to step 2

Step 2: Process “model.onnx”

  • In the On-device performance section, after a few minutes of process, I get the error.

Essentially the error presents as soon as i have uploaded the model.