3 ms·Demo: Decoupled Momentum Optimization – training neural networks in parallel3 points by sergiotapia 2y ago