Philipp Emanuel Weidmann

Note that the llama.cpp implementation of DRY has an additional parameter, `dry_penalty_last_n`, that allows controlling the length of the penalty context independently of the standard repetition penalty. I had originally...

I have added documentation for each parameter to the text-generation-webui frontend in this commit: https://github.com/oobabooga/text-generation-webui/pull/5677/commits/2585d07e2a0f84f86371b862b80b80daaf76f1d4 It might make sense to copy those documentation strings into this PR, perhaps in the...

Is `.` a commonly used notation for "or" then, or is that just a typo?

This has been brought up several times before and I have laid out my reason for not having the model name as a parameter. Namely, with the actual OpenAI API,...

Thank you for implementing this so quickly! I have submitted my review in the form of a pull request into this branch: #645.

@EricLBuehler Give me a few days to test and verify this, I will let you know once I'm sure. Looks fine at first glance though!

@EricLBuehler I'm having a hard time testing this properly because of #666.

@EricLBuehler I've played around some more and it seems the issue runs deeper than that. In fact, I've yet to find a model that actually works properly. Here's an example:...

@EricLBuehler I keep trying, but the bottom line is that inference is just failing for me with every combination. I'm going to provide as much information as I can to...

Thanks for the effort, but this is not the right way to broaden loader support. The right way is to add support for the text completion endpoint to those loaders...