Description
Describe the issue
Env:
- RK3576: Debian 12 aarch64, CPU only
- PC : Ubuntu20.04, x86
Ram
- 8GB
Description:
Hello,
- I am currently working with the onnxruntime-training demo, which I've successfully run on my PC and on Colab using Python, achieving satisfactory results. Now, I wish to deploy this project on my development board, a Rockchip RK3576. However, I've encountered an issue where I can't directly install the necessary packages via:
pip install onnxruntime
pip install onnxruntime-training
It appears that PyPI does not offer pre-compiled packages for my development board's environment. After some research, I found the following pre-compiled packages at this release page: https://github.com/Microsoft/onnxruntime/releases/tag/v1.19.2. I believe the files I need are:
-
onnxruntime-linux-aarch64-1.19.2.tgz
-
onnxruntime-training-linux-aarch64-1.19.2.tgz
Is my understanding correct?
-
As I am new to this, I'm unsure how to use these .tgz files with Python after downloading. Can the project only be invoked via C/C++, or is there a way to utilize these with Python?
-
Can I perform On-Device Training (CPU Only) using Python or C/C++ on Debian 12, arm64? Are there any example codes for ONNX Runtime training in C/C++? I have not found any relevant information so far, only Python code.
Thank you!
I look forward to your response, thank you!
Urgency
Yes, this issue is quite urgent for me as it affects my current work progress. I hope to receive a response as soon as possible so that I can continue advancing my project. Thank you for your understanding and support!
Target platform
aarch64
Build script
None
Error / output
None
Visual Studio Version
No response
GCC / Compiler Version
No response