David Pitt
David Pitt
Hi @lyyc199586 , thanks for opening this. You're right that this is an issue in torch 2.6, so I've opened #559 to fix this
Opened #563 to address this
I think making a fully stateless trainer would be a better use of time if we're going to change the Trainer interface significantly. What do you think? On Tue, Sep...
This sounds good to me On Wed, Sep 10, 2025 at 5:51 AM Jean Kossaifi ***@***.***> wrote: > *JeanKossaifi* left a comment (neuraloperator/neuraloperator#639) > > > I agree, it seems...
I think it's worth adding still as it's a nice simple demo. Will update soon
Hi @sjh97 , thanks for catching and opening this fix. Please remove the ipynb from your PR and place the test cases you'd like to verify into the `@pytest.mark.parametrize` statements...
One note on the correctness test: A relative tolerance of 5e-1 is really high. If the outputs do not pass a lower tolerance, e.g. 1e-5 or 1e-7, this means there...