Context Templates #36
Draft
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
I asked on the kobold horde channel about which llama 3 models actually use the instruct template. It seems like most do, but some do not, and one even uses ChatML (it has chat in the name). So, the problem is hard to overcome without building our own internal database of which models use which format. Which could be fun. But feels very out of scope for the goals of this project... |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
So far this adds Llama 3 instruct context template in a very hacky way. After implementing it and experimenting with trying to figure out which Horde models are available use it, I felt like it wasn't really worth it. Context templates is very much an advanced user feature that should, hopefully, go away, as LLMs progress. If the need doesn't go away, then I should consider implementing context templates in more robust way similar to Ollama/SillyTavern and not half-assing it like what this PR looks like currently.
Either way, I think I am going to hold off on this and not include it as part of 1.1. I also think I am going to revert some last minute changes to the default context template...