TensorNetwork icon indicating copy to clipboard operation
TensorNetwork copied to clipboard

Published 20 hours ago • verified publisher icon google

Tensor initialization with dtype and pytorch backend breaks

Open mganahl opened this issue 4 years ago • 8 comments

Tensor initialization methods for pytorch currently break for this type of initialization:

t = tn.random_uniform((2,2), dtype=np.float64, backend='pytorch')

(works for JAX and tensorflow) @alewis can you take a look?

mganahl avatar Sep 07 '20 09:09 mganahl