Removing threshold branch at inference while preserving weights ?

[charlesmindee]

Hi,

When you load the .h5 weights at inference time on the inference model, I don't really understand how you managed to save only the weights for the "inference model" (without threshold branch) while training ?

Thank you !