how can i solve it, please help me (kl630, optimize)

byeongju limbyeongju lim
in KL630 related discussion

Hello

I'm working on installing the yolox model in the kl630 product, and i have converted yolox_custom.pth to onnx, so if i proceed with the optimization work on the github of the kneron, the following picture will be obtained.

I'm contacting you like this because I don't know

error.png

how to solve the problem


My work environment is working on Windows 10 using wsl and docker:v0.19.0.

directory.png
docker.png

On the kneron inquiry page, i will leave a link to Google Drive containing the onnx file and py file because the upload capacity is up to 50mb

https://drive.google.com/file/d/1HEWqtku2tJ-F6JKkk5rnRmLC945Dxnf7/view?usp=sharing

Tagged:

Comments

  • byeongju limbyeongju lim

    I'm currently working on are RTX3050 (CUDA 11.1 / CUDNN 8.1.1 / PyTorch 1.9.0), and I'm working on this computer by converting the pth file that learned yolox to onnx. Is it only possible to use the tools in kneron's github?

  • Maria ChenMaria Chen

    Hi,

    According to the optimized onnx model, it seems like the correct input name for the onnx model is "images"

    image.png
    image.png

    Please change the "input" into "images"

    We also recommend you to write the preprocess function:

    image.png

    And process the images with the function:

    image.png


The discussion has been closed due to inactivity. To continue with the topic, please feel free to post a new discussion.

© Kneron, Inc.