You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm trying to use CrypTen's crypten.nn.from_pytorch() method to load YOLO5 from ultralytics. However, I have been running into some issues with ONNX functions Floor and Resize operations not being supported. I have seen other issues that discuss using the crypten.nn.module.py file to add additional functionality, but I'm unsure of how to add floor with the functions provided by the default cryptensor methods. I was considering adding methods to the mpc tensor structures themselves but I'm not exactly sure how to do this either, or if this is even the best approach to take.
Do you have any recommendations for me, and do you know how I should go about implementing a new cryptensor method if I wanted to do that (not just a module built on top of existing cryptensor methods)?
Thanks for the help!
The text was updated successfully, but these errors were encountered:
Hi,
I'm trying to use CrypTen's
crypten.nn.from_pytorch()
method to load YOLO5 from ultralytics. However, I have been running into some issues with ONNX functions Floor and Resize operations not being supported. I have seen other issues that discuss using the crypten.nn.module.py file to add additional functionality, but I'm unsure of how to add floor with the functions provided by the default cryptensor methods. I was considering adding methods to the mpc tensor structures themselves but I'm not exactly sure how to do this either, or if this is even the best approach to take.Do you have any recommendations for me, and do you know how I should go about implementing a new cryptensor method if I wanted to do that (not just a module built on top of existing cryptensor methods)?
Thanks for the help!
The text was updated successfully, but these errors were encountered: