llama-cpp-python
llama-cpp-python copied to clipboard
Add support for XTC and DRY samplers
Seems to work, but someone who knows this project better please check the order in apply_func().
This is a nice addition, as those parameters are present in llama.cpp but missing in the bindings. @abetlen
I would also love to see this merged!
Patiently waiting for the merge
Updated for llama.cpp changes
The sampler order was wrong, causing crashes when using XTC. Fixed now.
Hi, should dry_range be changed to dry_penalty_last_n to match llama.cpp?
Also, should the default value be changed from 0 to -1?
From llama.cpp/examples/main/README.md
--dry-penalty-last-n N: Set DRY penalty for the last n tokens (default: -1, 0 = disable, -1 = context size).
Good suggestions, thank you. Updated the PR.
@abetlen Hi, is it possible for you to merge this?
Edit: Please? :(
@abetlen Any chance of this getting a review?
@abetlen bump
@abetlen Could you please suggest what to do to move forward with this pull request?