Open utterances-bot opened 4 years ago
Hello Marie,
at the risk of blowing the scope of this course before it even starts:
I'm currently porting a simulation program (essentially a tensor network like in ML, but with dense matrices) in cupy
and dask
for
(a) running on GPU if the matrices fit in the GPU's memory
(b) on a single node with CPU only
(c) distributed across a number of nodes, if one node's memory in insuficcient.
It looks like PyTorch
is in some sense a frontend/wrapper for both cupy
and dask
. Do you know if that is correct? Or do you have any idea who might be able to help me with this?
Thanks, Philipp
PyTorch is not a front end of CuPy and Dask. There are similarities in the concepts, but it would be better described as an alternative.
I guess, if you wanted to run your simulation in PyTorch instead of CuPy and Dask, that should work too.
Installing Python and PyTorch · Machine-learning
https://westgrid-ml.netlify.app/school/pt-04-install.html