Skip to content

ggml sync #1082

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
leejet merged 2 commits into from
Dec 13, 2025
Merged

ggml sync #1082

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e3e8122
sync to last ggml sync with llama.cpp
catap Dec 12, 2025
a9ee71e
fix api
catap Dec 13, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ggml
Open in desktop
Submodule ggml updated 251 files
4 changes: 4 additions & 0 deletions ggml_extend.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1282,13 +1282,17 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_ext_attention_ext(struct ggml_context
}

if (mask_in != nullptr) {
// the need for padding got removed in ggml 4767bda
// ensure we can still use the old version for now
#ifdef GGML_KQ_MASK_PAD
int mask_pad = 0;
if (mask_in->ne[1] % GGML_KQ_MASK_PAD != 0) {
mask_pad = GGML_PAD(L_q, GGML_KQ_MASK_PAD) - mask_in->ne[1];
}
if (mask_pad > 0) {
mask_in = ggml_pad(ctx, mask_in, 0, mask_pad, 0, 0);
}
#endif
mask_in = ggml_cast(ctx, mask_in, GGML_TYPE_F16);

@wbruna wbruna Dec 13, 2025

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And what's the reason for simply removing the padding here?

@catap catap Dec 13, 2025

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Because it was simple removed at ggml :)

See: ggml-org/llama.cpp#17910

@wbruna wbruna Dec 13, 2025

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, I see.

But you are assuming we'll only need to build against the new version, once it gets updated. Most of the time it's inevitable, due to an incompatible API, but being able to alternate between versions is very useful to help tracking down bugs introduced by the version change. And in this case, we can. I suggest doing something like this:

diff --git a/ggml_extend.hpp b/ggml_extend.hpp
index 92dd3b8..1d32f0e 100644
--- a/ggml_extend.hpp
+++ b/ggml_extend.hpp
@@ -1268,6 +1268,9 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_ext_attention_ext(struct ggml_context
         }
 
         if (mask_in != nullptr) {
+            // the need for padding got removed in ggml 4767bda
+            // ensure we can still use the old version for now
+            #ifdef GGML_KQ_MASK_PAD
             int mask_pad = 0;
             if (mask_in->ne[1] % GGML_KQ_MASK_PAD != 0) {
                 mask_pad = GGML_PAD(L_q, GGML_KQ_MASK_PAD) - mask_in->ne[1];
@@ -1275,6 +1278,7 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_ext_attention_ext(struct ggml_context
             if (mask_pad > 0) {
                 mask_in = ggml_pad(ctx, mask_in, 0, mask_pad, 0, 0);
             }
+            #endif
             mask_in = ggml_cast(ctx, mask_in, GGML_TYPE_F16);
         }

}

Expand Down
Loading