One-Step Forward and Backtrack: Overcoming Zig-Zagging in Loss-Aware Quantization Training

Open in new window