llama.cpp
convert : use n_groups instead of hardcoded values in reshape
#18929
Merged
Merged
convert : use n_groups instead of hardcoded values in reshape #18929
danbev
merged
2 commits into
ggml-org:master
from
danbev:convert-nemotron-nano-v3
convert : use n_groups instead of hardcoded values in reshape
dac1fea8
CISC
approved these changes
on 2026-01-19
use self.n_group instead of self.hparams["n_groups"]
535f8e34
Login to write a write a comment.
Login via GitHub