Bryce Meyer

@mntss It is completely certain that the issue is implementation inaccuracy. This topic has been discussed a lot over the last few months. If you are curious about the details,...

@jqhoogland Thanks for bringing this up. I think @jbloomAus is already aware of this, but I am in the middle of setting up some benchmarking tools for TransformerLens https://github.com/TransformerLensOrg/TransformerLens/tree/benchmark-utlitities. I...

We are about ready to put up a release where all einsum implementations have been replaced with standard PyTorch functions. Issues like this will no longer be an issue, so...

@HenryCai11 Do you need any help getting started with this? If you want to put the demo together, I would be happy to walk you through a couple steps in...

Someone has inadvertently experimented with this, and it does seem like it is not going to be a trivial process. I don't know if it is going to be the...

@jettjaniak It may have been a different dependency that I am thinking of. I will try to update it to 0.18 now, and see if we have any major issues...

Alright! I think I was recalling a different tool when I posted my first comment. This upgrade was pretty trivial. It did require an update to dependencies for torch, so...

Well it was actually mypy that was the issue for the dependency update! That is still an issue, so I am reopening this one until those mypy changes are resolved.

There are a lot of issues in this pr due to dependency bumping. None of that has anything to do with what has been done here, but there are general...

The failure is due to these tests now triggering http requests on gated models. GitHub does allow for two buckets of variables, and we could put the hf token into...