Kai Londenberg
Kai Londenberg
Just as a quick comment. If the goal is to keep the code as readable and intuitive as possible, I would think it makes sense to branch this out. It...
You are right that there is no paper about it. It's a pretty trivial nonlinearity which I came up with. Here is why I consider it useful: 1.) it can...
> Add a test that checks this works with dynamic shapes. Did that..
Closing. The PR stack is being restructured here: https://github.com/pytorch/pytorch/pull/121492
Closing. The PR stack is being restructured here: https://github.com/pytorch/pytorch/pull/121492
> @kadeng , does torch have this need? Not at this moment, but the argument about improving Cudagraph reuseability appears compelling.
Tried that, both do not make a difference. On a sidenote, I had to update the linked gist, since I noticed that the code formatter I used directly before pasting...
No, but I did not try on the latest Cutlass version.
If this example is changed such that the loaded auxiliary operand is of the same shape but not the same (pointer) as operand A, the error does not happen. So...