search

rckirby / torchfire

Embeds Firedrake functionality into PyTorch.
MIT License
1 stars 0 forks source link

Data types #2

Open rckirby opened 2 years ago

rckirby commented 2 years ago

Currently, fecr seems to require float64 data types. Hence, we're hard-wiring that into our stack. If fecr can be generalized, we should accommodate that on our end.

rckirby commented 2 years ago

Looks like this was generalized here: https://github.com/IvanYashchuk/fecr/pull/15/files

So @Re-GeorX we should remove that restriction in our code we can probably either ask a Function what its dat's datatype is or read something globally set in PyOP2.

rckirby commented 1 year ago

@Re-GeorX is this stll live? We should close the issue if it's been fully resolved.

Re-GeorX commented 1 year ago

This is still live. I think there's a little more work to do in fecr to work with datatypes other than float64. Until fecr has not everything fixed I will keep imposing the cast to float64. I probably will open a new issue in Ivan's repo for this.