Closed michaeljohns2 closed 3 years ago
Tested on both tf 1.15 and 2.2 and logic worked.
This issue is fixed in benepar 0.1.3.
Starting with benepar 0.2.0a0, I've also switched to not using tensorflow at all. My original hope of having this package remain usable without any ongoing maintenance has sadly proven to be inadequate. Unfortunately modern ML code seems to have a short shelf-life -- meanwhile software like the EVALB program checked into this repo remains usable despite no changes over the course of a decade or more. With v0.2.0, I'm switching to a new approach where the code has more dependencies, but it's also easier to maintain because I'm using these same dependencies for my research and I have to keep up with software changes anyway. (This is different from tensorflow, which I don't use myself because I find pytorch and jax to better serve my research needs)
Updated for tensorflow v1 or v2 compatibility.