作者您好,请问model_common_train.py的931行代码
filter_curr_stroke_image_soft = tf.multiply(tf.subtract(1.0, curr_state_soft), curr_stroke_image_large)
这里是不是应该改为
filter_curr_stroke_image_soft = tf.multiply(curr_state_soft, curr_stroke_image_large)
不需要tf.subtract(1.0, curr_state_soft)这个一步操作