fix: llama3.1 rope_freqs not respecting custom head_dim #9141
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
This comment was marked as spam.
This comment was marked as spam.
Sorry, something went wrong.
ggerganov
approved these changes
Aug 23, 2024
4 tasks
|
Any ETA on merging this? It seems fully functional |
|
PTAL @ggerganov |
|
ily @ggerganov |
|
@LostRuins CHOP CHOP |
|
Confirmed working fine for me, tested with a fresh quant. Note that all Llama-4b-minitron GGUF models converted prior to this PR should be reconverted. |
dsx1986
pushed a commit
to dsx1986/llama.cpp
that referenced
this pull request
Oct 29, 2024
…org#9141) * fix: llama3.1 rope_freqs not respecting custom head_dim * fix: use potential head_dim for Exaone
arthw
pushed a commit
to arthw/llama.cpp
that referenced
this pull request
Nov 15, 2024
…org#9141) * fix: llama3.1 rope_freqs not respecting custom head_dim * fix: use potential head_dim for Exaone
arthw
pushed a commit
to arthw/llama.cpp
that referenced
this pull request
Nov 18, 2024
…org#9141) * fix: llama3.1 rope_freqs not respecting custom head_dim * fix: use potential head_dim for Exaone
Nexesenex
pushed a commit
to Nexesenex/croco.cpp
that referenced
this pull request
Feb 25, 2025
…org#9141) * fix: llama3.1 rope_freqs not respecting custom head_dim * fix: use potential head_dim for Exaone
Nexesenex
pushed a commit
to Nexesenex/croco.cpp
that referenced
this pull request
Feb 25, 2025
…org#9141) * fix: llama3.1 rope_freqs not respecting custom head_dim * fix: use potential head_dim for Exaone
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Outdated suggestions cannot be applied.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.
The Llama 3.1 pruned Minitron models utilize the llama3 rope_scaling and has a custom head_dim specified in the config.
This changes the conversion code to use the custom head_dim for llama models if it is specified, otherwise falling back to the old calculation.
n_rot seems to be present no matter the model, as it still works on both regular llama-3 and llama-3.1 with this change.
This fixes #9060