Skip to content

Pull requests: ggml-org/ggml

New pull request New
51 Open 855 Closed
51 Open 855 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

simple : temporary change to trigger Vulkan crash on macOS
#1623 opened Sep 9, 2026 by ggerganov Member • Draft
1
Fix builds on s390x
#1620 opened Sep 6, 2026 by WhyNotHugo Loading…
1
vulkan: only request 32-wide FA subgroups where size control can grant them
#1619 opened Sep 4, 2026 by thewh1teagle Contributor Loading…
vulkan: fall back to vkGetDeviceQueue when vkGetDeviceQueue2 returns null
#1616 opened Sep 4, 2026 by thewh1teagle Contributor Loading…
1
vulkan: load the Vulkan loader at runtime instead of linking it
#1614 opened Sep 4, 2026 by thewh1teagle Contributor Loading…
ggml-opencl: guard F16 KQ/KQV path with use_adreno_kernels (fix SD3.5 NaN)
#1613 opened Sep 2, 2026 by Hmission Loading…
ggml-opencl: Mali tiled GEMM half-precision variant (2.86x on Mali-G925)
#1612 opened Sep 2, 2026 by Hmission Loading…
sgemm: handle m % 4 != 0 instead of declining the whole matmul
#1604 opened Aug 21, 2026 by femelo Loading…
sgemm: write the operand addresses so GCC will strength-reduce them on aarch64
#1603 opened Aug 21, 2026 by femelo Loading…
sgemm: don't pick a register tile GCC will spill on aarch64
#1602 opened Aug 21, 2026 by femelo Loading…
Fix Qualcomm HTP and Vulkan execution on Windows ARM64
#1601 opened Aug 21, 2026 by roozbehid Loading…
Handle F32 x F16 src1 in CUDA binary broadcast ops
#1591 opened Aug 20, 2026 by Bonenk Loading…
Add WIN_PART/WIN_UNPART + head-size-32 attention for SAM3
#1590 opened Aug 18, 2026 by greenjava Loading…
fix(metal): do not latch has_error on transient GPU failures
#1585 opened Aug 16, 2026 by olxqnz5et7g2rv Loading…
2 tasks
1
Implement GGML_OP_IM2COL_3D for Metal backend
#1574 opened Jul 27, 2026 by venomega Loading…
Gate RVV code on the target extension
#1571 opened Jul 19, 2026 by catap Loading…
1
opencl: add Adreno xmem SDPA path
#1562 opened Jul 11, 2026 by happyyzy Contributor Loading…
opencl: fix Q4_0 SOA rows and Adreno paths
#1561 opened Jul 11, 2026 by happyyzy Contributor Loading…
3
sycl: fix soft_max_f32 max reduction
#1534 opened Jun 10, 2026 by someoneinjd Contributor Loading…
2
metal : graceful failure instead of GGML_ABORT
#1527 opened Jun 3, 2026 by fiorelorenzo Loading…
tests : add OLMoE-sized K-quant shapes to test_mul_mat_id (ref #1506)
#1518 opened May 28, 2026 by devYRPauli Contributor Loading…
3
Fix build and test running for Windows MSVC
#1512 opened May 26, 2026 by MaxwellGengYF Loading…
docs(gguf): extend Encoding slot to support percentage-mixed recipes
#1489 opened May 13, 2026 by mishig25 Contributor • Draft
1
ggml-cpu/riscv: gate cpu-riscv64 backend on Zv* sub-extensions
#1475 opened May 2, 2026 by mikey Loading…
opencl: add Adreno xmem attention fast path
#1461 opened Apr 19, 2026 by happyyzy Contributor Loading…
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.