How to decide the margin in Triplet Loss to be used in the training of a Siamese Network

vision
1

I am not sure how much margin should i keep in my Triplet Loss. My model is not training.
Here is my implementation of the Siamese Network