Getting inconsistent outputs for same inputs when encoding the images
#4
by
f10409
- opened
I wonder if we need to put the model into eval mode in the load_model function in medimageinsightmodel.py.
Thanks!
I also noticed this issue. In example.py, after modified the code and test several private use cases, I get consistent outputs for same inputs and the zero-shot classification result shows that the modified content seems to be correct.