llama.cpp
Fix kq_scale for the attention layers of PLaMo2
#14892

Merged

Fix kq_scale for the attention layers of PLaMo2 #14892

CISC merged 7 commits into ggml-org:master from mitmul:mitmul/fix-build-attn-scale-plamo2

Fix dimensions for expand

7baf4fd1

Change dimensions to copy states to cache

e39bc092

Fix the default value for plamo2 conversion

bd4d2e1c

Fix scale given to build_attn

c475203c

github-actions added python

CISC approved these changes on 2025-07-26

CISC commented on 2025-07-26

Update src/llama-model.cpp

75f0a0d7

Update src/llama-model.cpp

429639d0

Update src/llama-model.cpp

60a705de

CISC merged 1dc9614e into master 139 days ago

mitmul deleted the mitmul/fix-build-attn-scale-plamo2 branch 139 days ago

Reviewers

CISC

Assignees

No one assigned

Labels

python

Milestone

No milestone