Add epsilon and eta sampling #475
Merged
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
|
@henk717 this looks good to me, have you tested yet for a merge? |
g4rg
added a commit
to g4rg/aphrodite-engine
that referenced
this pull request
Oct 3, 2023
|
Have not been able to review it yet (See discord for details), its the first one on my list once I am able to review. |
AlpinDale
pushed a commit
to aphrodite-engine/aphrodite-engine
that referenced
this pull request
Oct 4, 2023
* KoboldAI endpoint * KAI: fix loading, spec defaults, error handling * ignore use_default_badwordsid * prefix model name * KAI: fix incorrect JSONResponse usage * KAI: reduce schema * KAI: implement kobolcpp style streaming * KAI: use list comprehension * KAI: lint * KAI: add epilon+eta cutoff parameters according to henk717/KoboldAI#475
|
Can't reproduce the issue I had with it anymore, so ill merge it and hopefully we don't break userspace. |
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Outdated suggestions cannot be applied.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.
I have implemented epsilon and eta sampling from https://arxiv.org/abs/2210.15191 .
In the UI I opted to expose it in units of 1e-4, to be similar to ooba.
Tested both pytorch/gpu and TPU(jax_static specifically) and everything seems to work fine.