Open HannesStark opened 1 year ago
Hi Hannes, unfortunately, this is the ugliest hack you have ever seen. This is a uniform distribution that tolerates samples sloshing slightly over the boundaries due to small numerical errors in the flow. The implementation is gruesome though.
Hi and thanks for the library!
What is the purpose of SloppyUniform instead of torch.distributions.uniform.Uniform and why might it be causing my debugger to no longer stop at breakpoints? https://github.com/noegroup/bgflow/blob/fbba56fac3eb88f6825d2bd4f745ee75ae9715e1/bgflow/distribution/distributions.py#L81
Thanks for any insights!