I encountered an error while trying to convert a bie file to a nef file.
Hello, I am currently working on mmclassification.
But, I encountered an error while trying to convert a .bie file to a .nef file.
I'm using kneron toolchain version 0.21.0, and I've tried using different versions without success. I've also attempted to change the ONNX and PyTorch versions, but the results remain the same.
What steps should I take to resolve this issue? Your help would be greatly appreciated. Thank you
The code used for the conversion of bie to nef is as follows:
Tagged:
The discussion has been closed due to inactivity. To continue with the topic, please feel free to post a new discussion.
Comments
Hi JungTae,
Could you provide us with your results for your IP Evaluator and all of your files, including your onnx file? If all the previous steps worked smoothly, this might be a compiler issue. Thank you!
The error message says, "Input onnx is missing." Could you please explain what this means?
It's an ONNX file and pth file. And below is the successful message of ONNX conversion.
Below are the test results of the pth file.
Below are the test results of the model converted to ONNX
And, Could you please provide more details on what you mean by 'IP evaluator'? I'm not familiar with it. Thank you for your assistance.
Hi JungTae,
Sorry for the late reply. The error message "Input onnx is missing" means that your input is missing the onnx file. Please make sure you're including an onnx file as an input.
Thank you for providing the onnx files. This might not be a compiler issue, since I assumed that your toolchain generated bie file properly, but I didn't see any bie files attached, and when we tried to convert the model to nef file, the nef file did generate.
The sample code above has a random image for input_images. Please feel free to use more images to keep the model accuracy.
An example on converting ONNX to nef: 1. Toolchain Manual Overview - Document Center (kneron.com)
IP Evaluator is a tool that is mainly used to check your model's performance estimation and whether your model is supported by the toolchain.
Information on IP Evaluator can be found here:
3. Floating-Point Model Preparation - Document Center (kneron.com)
Script Tools - Document Center (kneron.com)