Possibility to wrap scikit-learn functions in Variable and run in gpu

1

Hi everyone, I was wondering if i’ts possible to wrap scikit-learn’s functions in something like let’s say Variable and at run time make them execute on gpu instead of cpu? Does anyone know of anything similar?

2

Hi,

scikit-learn functions are implemented for cpu, so it is not possible to run it on gpu.
You would need to use gpu-implemented functions.

3

Thanks! I know that scikit is cpu but I thought maybe there was some dirty hack or trick to make that work on gpu without too much hustle. There’s h2o4gpu which is trying to translate scikit to gpu based algos but still is very early and hasn’t adopted most of the algos yet.

4

Which algorithm do you want to use?

5

I’m trying to use isomap.

6

Ho, I’m afraid I’m not aware of anything like that done with pytorch Tensors/on gpu.

7

Thanks! Apparently there is https://github.com/h2oai/h2o4gpu but still needs a lot of work to be done.

8

@kirk86 Have you found anything? How did you go about doing it?

9

Apologies for the late reply. Apparently not yet. I haven’t had the time to look into it properly.