Stephen Roller
Stephen Roller
This is good, all of this should be turned into a docs page. Thanks for asking these great questiosn.
Ah, the BlenderBot3 and 9 models have `--model-parallel true` which doesn't mix with `multiprocessing_train`, sorry. Those are trained as is, because they're so big lol.
(Thinking about your proposal, it's interesting)
Dexter, please embed multilines when you quote code, the issues are difficult to read without this.
I don't think a default variant should be xlm, it should be aiayn if anything, but really we should just make None behave the same as aiayn.
> It would be nice to have better installation instructions and specific package dependencies listed in the [hallucination project](https://parl.ai/projects/hallucination/). Right now it's an ocean of upgrading and downgrading packages until...
Good catch, thanks
Now that we have some data caching, I think this is actually much more impactful, @dianaglzrico
Wasn't aware of it. It looks promising. @shanemoon, you interested?
Can you describe when it happens?