deep-learning sam optimizer pytorch ranger loss-functions chebyshev lookahead nero adabound learning-rate-scheduling radam diffgrad gradient-centralization adamp adabelief madgrad adamd adan adai

transformer wgan-gp pytorch-implementation madgrad