Paul

> > @paul0403 Just realized the change of the name of slice to tuple caused test failures, that should also need the pennylane to change all the occurrence of `..._slice`...

Soft approval from me, pending @mehrdad2m 's conversations being resolved

I think this is because the `static_argnum` tracks the argument indices at the call, not at the definition, though this is just a hunch and might be completely wrong

Hi @AniketDalvi , do you have a more specific question in mind? I would love to give out more pointers and resolve any confusion you may have!

> Hi! So for installation - it says to download the PyPI wheel. Which version should I be downloading the wheel for? `pip` says there are 3 available versions -...

> Okay when trying to run the `make frontend` command with the new wheel, I get this error - `error: command '/usr/bin/g++' failed with exit code 1`. I have tried...

(If wheels are too complicated I recommend just building from source.)

> Yup installed all the required packages, but the error persists. I am now just going to build from source instead. To avoid all package version issues, I also recommend...

> Hi! So from my initial analysis, I traced the issue down to this check that throws an exception - > > https://github.com/PennyLaneAI/catalyst/blob/75dc517a6e2f2583da890b9e8198241933a7aef5/frontend/catalyst/tracing/type_signatures.py#L120 > > . > It appears that...

> * a very large amount of constant data is embedded into the IR during the hlo lowering process This big tensor is generated during [scattering](https://github.com/PennyLaneAI/catalyst/blob/c0803e588c0abe50b6c4cc725b993c3b525dac87/mlir/lib/Catalyst/Transforms/ScatterPatterns.cpp#L261), and is followed by...