However, this first creates CPU tensor, and THEN transfers it to GPU… this is really slow. ... and makes only a portion of the data available to that GPU.
確定! 回上一頁