When I use numpy to process the tensor array generated by tensorflow, does it generate a new numpy array in memory or directly use the passed in tensor

[jun_yin]

(topic deleted by author)