Skip to content

fix: sched split-input overflow on Gemma 4 E-series partial offload #46

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
worthant merged 3 commits into from
Jul 24, 2026
Merged

fix: sched split-input overflow on Gemma 4 E-series partial offload #46

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d7b42ab
Merge pull request #45 from AtomicBot-ai/dev
worthant Jul 23, 2026
d4db42d
release: b10018-1.1.1
worthant Jul 23, 2026
8fcf218
🔧 fix(ggml): raise sched split-input limit for per-layer-input models
worthant Jul 24, 2026
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 TURBOQUANT_VERSION
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1 +1 @@
b10018-1.1.0
b10018-1.1.1
6 changes: 5 additions & 1 deletion ggml/src/ggml-backend.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -754,7 +754,11 @@ static bool ggml_is_view_op(enum ggml_op op) {
#endif

#ifndef GGML_SCHED_MAX_SPLIT_INPUTS
#define GGML_SCHED_MAX_SPLIT_INPUTS 30
// 30 upstream. Gemma 4 E-series feeds a per-layer embedding input into every
// block; with partial offload (small-VRAM Vulkan + CPU) each of them becomes
// a split input, so a 30+-layer model trips the assert at split creation.
// 128 covers the deepest current target (31B dense) with headroom.
#define GGML_SCHED_MAX_SPLIT_INPUTS 128
#endif

#ifndef GGML_SCHED_MAX_COPIES
Expand Down
Loading