Hi, While training the network on my own dataset, the losses ires_disp_loss0, ires_disp_loss1 and ires_disp_loss2 looks same. Is this seem reasonable ? 