Guillaume Lagrange

> I added it to see if there was any value. But I haven't seen useful results yet. Eventually maybe. I'll keep trying it out. It doesn't seem to hurt....

We can probably close this in favor of #3254 @catch-twenty-two ?

No worries! Didn't want to close without checking if you had other plans in mind for this

Using the `Backend::seed(seed)` method should do the trick. ```rust type B = NdArray; let device = NdArrayDevice::default(); B::seed(42); let conv2d = Conv2dConfig::new([1, 1], [1, 1]).init::(&device); dbg!(conv2d.weight.val().into_data().to_vec::().unwrap()); ``` It's used across...

We should use the tensor data equality methods defined, not `i_output.equal(i_expected).all().into_scalar()`. Linking #3134 for reference as it points to the bool underlying data type discrepancy with wgpu backend. Ofc this...

Ahh yes the "Fixes [issue number]" pattern in a PR automatically closes the issue when the PR merges 😅 Reopening for the other related issues mentioned in the comments.

No worries! In this case I think it's fine to have a single issue, but you can just link it without closing it as long as you don't have "fixes/closes/resolves...

Also, ignore the macos CI failing.. we're having issues with the space on the runners since yesterday 🥲

Hmmm looks like the model used in the examples actually uses an id format that predates the currently covered backward compat (parameter ids look like uuids `30b82c23-788d-4d63-a743-ada258d5f13c`). I think I'll...

Yeah totally forgot the param ids used to be 128-bit UUIDs before it was changed in #1912, and the `mnist-inference-web` model I was trying to convert was generated before the...