With the stalling of Haskell Tensorflow (i.e, Stack only without recent support), we’re left with HaskTorch as a full-featured ML library for Haskell. However, the documentation suggests that use of 0.1 is discouraged, and 0.2 is still in progress. The last update on the Github was in May.
Is there a roadmap to getting HaskTorch usable? Or should we be more looking at Grenade, which is still LTS, but requires dependent types?