Zoom Download original image Report abuse

Training dynamics under different $\gamma$ and lr.