使用add_position_encoding在GPU上训练报错
Created by: xprincehf
使用fluid.layers.add_position_encoding的时候,在CPU上可以正常训练,但是用GPU会报错 op add_position_encoding does not have kernel for data_type[float]
类似问题的issue有这些,该问题是否是同样bug #14378 (closed) #12255 (closed) #14566 (closed)