Slight quantization improvement for Q4_K and Q5_K #5361
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
ggerganov
approved these changes
Feb 6, 2024
|
@ikawrakow could you add some PR links to recent quantization changes in the readme? here are some added a good while back: https://github.com/ggerganov/llama.cpp#quantization |
See #5366 |
jordankanter
pushed a commit
to jordankanter/llama.cpp
that referenced
this pull request
Mar 13, 2024
* Q4_K: slightly better quantization * Q5_K: slightly better quantization --------- Co-authored-by: Iwan Kawrakow <[email protected]>
hodlen
pushed a commit
to hodlen/llama.cpp
that referenced
this pull request
Apr 1, 2024
* Q4_K: slightly better quantization * Q5_K: slightly better quantization --------- Co-authored-by: Iwan Kawrakow <[email protected]>
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Outdated suggestions cannot be applied.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.
Nothing earth-shattering, but I noticed that I have not brought over all changes I made to
Q4_KandQ5_Kquantization in my repo, so here it is. We get small PPL improvements. E.g., for Mistral-7B with context of 512 and imatrix fromwiki.traiin.raw(QErroris defined as PPL(Q)/PPL(fp16)-1, where PPL(fp16) = 5.6924 for Mistral-7B):