model: add missing build_cvec to plamo2#21444
model: add missing build_cvec to plamo2#21444nisparks wants to merge 1 commit intoggml-org:masterfrom
Conversation
PR ggml-org#20653 added control vector support to models that were missing it, but the plamo2 change was incomplete — it added the comment ('// input for next layer') but missed the actual build_cvec() call. Compare the ggml-org#20653 diffs for the two sibling models: plamo3.cpp: +build_cvec(cur, il) +cb(cur, "l_out", il) +// input for next layer plamo2.cpp: +// input for next layer (build_cvec omitted) Without this call, --control-vector is silently ignored on PLaMo-2. Normal inference is unaffected since build_cvec is a no-op when no control vector is loaded. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
|
Hi @nisparks, thanks for your contribution! Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:
Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below. |
|
Not missed, but rather purposefully left out, see #20653 (review) Also, please read and conform to our contribution guidelines in the future. |
Summary
PLaMo-2 is missing the
build_cvec()call in its graph builder, causing--control-vectorto be silently ignored.Root Cause
PR #20653 ("model: add control vector support where missing") touched both
plamo2.cppandplamo3.cpp, but the plamo2 change was incomplete.Comparing the diffs from #20653:
plamo3.cpp — correct (3 lines added):
cur = build_cvec(cur, il); cb(cur, "l_out", il); // input for next layerplamo2.cpp — incomplete (only the comment was added):
// input for next layerThe actual
build_cvec()andcb()calls were missed.Fix
Adds the missing two lines, matching the pattern in
plamo3.cppand all other 100+ model implementations:cur = build_cvec(cur, il); cb(cur, "l_out", il);Impact
--control-vectorsilently does nothing on PLaMo-2build_cvecis a no-op when no control vector is loaded, so normal inference is completely unaffected