]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
feat: Add conversion support in GraniteHybrid for non-hybrid (all attn) (#16177)
authorGabe Goodhart <redacted>
Mon, 22 Sep 2025 18:40:10 +0000 (12:40 -0600)
committerGitHub <redacted>
Mon, 22 Sep 2025 18:40:10 +0000 (20:40 +0200)
commit1d0125bcf1cbd7195ad0faf826a20bc7cec7d3f4
tree956d26a0542d152616c9e208920b327192daec44
parent351f3da39c85f59d581fc184f09283da7f099a3b
feat: Add conversion support in GraniteHybrid for non-hybrid (all attn) (#16177)

This is a configuration of the hparams in the GraniteHybrid architecture
that devolves to the Granite (or GraniteMoe) architecture (ie Granite 3.x).
It may be used for some models in the Granite 4 family with the
GraniteHybrid architecture acting as a superset arch. Rather than support
it directly in the c++ graph, we simply coerce the architecture flag back
to the correct "granite" or "granitemoe" architecture.

Branch: gabe-l-hart/GraniteNonHybridConversion

Signed-off-by: Gabe Goodhart <redacted>
Co-authored-by: Sigbjørn Skjæret <redacted>
convert_hf_to_gguf.py