Closed Kelvinson closed 5 years ago
A not-thorough migration to support pytorch>= 0.4, solve issue 12, includes:
As @st2yang pointed out here, it can still use the provided trained weights.
A more thorough migration is on my branch, but it needs training from scratch and there is nan error.
Note the both of two branches only work for V-REP 3.5. A hack around to adapt V3.6 is welcome. Have fun!
A not-thorough migration to support pytorch>= 0.4, solve issue 12, includes:
As @st2yang pointed out here, it can still use the provided trained weights.
A more thorough migration is on my branch, but it needs training from scratch and there is nan error.
Note the both of two branches only work for V-REP 3.5. A hack around to adapt V3.6 is welcome. Have fun!