Distilling Knowledge from a image classification model with sigmoid function and binary cross entropy #213
Unanswered
publioelon
asked this question in
Q&A
Replies: 1 comment
-
Hi @publioelon Let me summarize your setting:
To apply conventional KD in PyTorch, you would use output from the final linear layer without softmax/sigmoid as CrossEntropyLoss and KLDivLoss are designed to use them. Besides that, which Dataset class in PyTorch are you using for your custom dataset? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello, I'd like to ask if it's possible to use this framework to use a custom trained model (input shape 128x160) and two classes fever and healthy to a smaller model, given that it used binary cross-entropy for loss computation and sigmoid function for classification? I need to compress the model size and switch to softmax classification due to layer compatibility with hardware accelerator (Edge TPU google coral accelerator).
Beta Was this translation helpful? Give feedback.
All reactions