i using keras, , loss funtion below. shape of y_true , y_pred suppose (,15,15,8,5). last 5 x,y,w,h,conf of box. hope can me. thanks! the output while training like: 8/1666 [..............................] - eta: 1662s - loss: nan - acc: 0.8994 the loss function is: def custom_loss(y_true, y_pred): ### adjust predictions # adjust x , y pred_box_xy = tf.sigmoid(y_pred[:, :, :, :, :2]) # adjust w , h pred_box_wh = tf.exp(y_pred[:, :, :, :, 2:4]) * np.reshape(anchors, [1, 1, 1, box, 2]) pred_box_wh = tf.sqrt(pred_box_wh / np.reshape([float(s_grid), float(s_grid)], [1, 1, 1, 1, 2])) # adjust confidence pred_box_conf = tf.expand_dims(tf.sigmoid(y_pred[:, :, :, :, 4]), -1) y_pred = tf.concat([pred_box_xy, pred_box_wh, pred_box_conf], 4) ### adjust ground truth # adjust x , y center_xy = 0.5 * (y_true[:, :, :, :, 0:2] + y_true[:, :, :, :, 2:4]) center_xy = center_xy / np.reshape([(float(w_image) / s_grid), (float(w_image) / s_grid)], [1, 1, 1, 1, 2]) true_box_xy = center_xy - t...