llama.cpp
models : fix the attn_factor for mistral3 graphs + improve consistency
#17945
Merged

models : fix the attn_factor for mistral3 graphs + improve consistency #17945

ggerganov merged 8 commits into master from gg/mistral-fix-attn-factor
ggerganov
ggerganov models : fix the attn_factor for mistral3 graphs
1df2e908
github-actions github-actions added model
ngxson
ngxson approved these changes on 2025-12-11
ggerganov cont : rework attn_factor correction logic
59b9e36f
ggerganov cont : make deepseek2 consistent
45930c97
ggerganov ggerganov force pushed from 0ca55b61 to 45930c97 188 days ago
ggerganov cont : add TODO
45875df2
ggerganov ggerganov marked this pull request as ready for review 188 days ago
ggerganov ggerganov requested a review from CISC CISC 188 days ago
ggerganov ggerganov requested a review from ngxson ngxson 188 days ago
ggerganov
ggerganov commented on 2025-12-12
github-actions github-actions added python
ggerganov cont : special-case DSv2
06eb8e86
ggerganov ggerganov changed the title models : fix the attn_factor for mistral3 graphs models : fix the attn_factor for mistral3 graphs + improve consistency 188 days ago
ggerganov cont : revert Mistral 3 Large changes
01b77b57
ngxson
ngxson commented on 2025-12-12
ggerganov cont : fix DS2 to use the original attn_factor
7320a2dc
ngxson
ngxson approved these changes on 2025-12-12
ggerganov
ggerganov commented on 2025-12-12
ggerganov cont : minor comments [no ci]
d6477e14
ggerganov ggerganov merged 7bed317f into master 188 days ago
ggerganov ggerganov deleted the gg/mistral-fix-attn-factor branch 188 days ago
Nindaleth

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone