forked from halo-box/llama.cpp
-
Notifications
You must be signed in to change notification settings - Fork 14
Pull requests: halo-box/strix-llama.cpp
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
cuda: enable the BF16 WMMA matmul path only for qwen4exp
CUDA
ggml
#75
opened Sep 17, 2026 by
dzannotti
Collaborator
Loading…
fix: guard the qwen4exp PLE prefetch on the model architecture (SIGSEGV at ubatch >= 4096)
model
#74
opened Sep 17, 2026 by
dzannotti
Collaborator
Loading…
trace : add DeepSeek V4.1 cross-runtime correctness harness
build
documentation
Improvements or additions to documentation
server
testing
#71
opened Sep 17, 2026 by
ajaxdude
Member
•
2/2
Loading…
deepseek41 : add core DeepSeek V4.1 runtime (arch, loader, Engram, expert store, memory admission, watchdog)
build
conversion
documentation
Improvements or additions to documentation
ggml
model
testing
#70
opened Sep 17, 2026 by
ajaxdude
Member
•
1/2
Loading…
4 tasks done
spec : cost-aware MTP draft length, MTP draft vocabulary subset, flat get_rows and shape-aware graph key
CUDA
ggml
model
testing
#58
opened Sep 14, 2026 by
SimonTeixidor
Loading…
llama : qwen35/qwen35moe GDN graph - 2-D output projection, in-place recurrent state, masked h_nextn fix
model
testing
#57
opened Sep 14, 2026 by
SimonTeixidor
Loading…
HIP : RDNA3.5 MMVQ/MMQ kernels and per-type dispatch thresholds for quantised matmul
CUDA
ggml
#56
opened Sep 14, 2026 by
SimonTeixidor
Loading…
cuda: ROCmFPx tensor types + Q4_0_ROCMFP4_FAST MMQ tiles on HIP (gfx1151)
CUDA
ggml
#40
opened Sep 10, 2026 by
baraxnaxgaming-commits
Loading…
DRAFT: ggml: optimize Flash Next inference paths
CUDA
ggml
testing
Vulkan
#39
opened Sep 10, 2026 by
gaetan-puleo
Collaborator
•
Draft
ggml : backport persistent view initialization after buffer-size splits
ggml
testing
#37
opened Sep 8, 2026 by
arc-uri-el
Loading…
server : preserve RAM-cached branches sharing a large prompt prefix
server
#36
opened Sep 8, 2026 by
arc-uri-el
Loading…
Fix/rocmfpx encoder review
ggml
testing
Vulkan
#31
opened Sep 7, 2026 by
LaurentZuijdwijk
Member
Loading…
vulkan: fix ROCmFPx mat-vec cost at batch 3-8 on rocmfpx/wholesale-reference
#7
opened Aug 29, 2026 by
LaurentZuijdwijk
Member
•
Draft
ProTip!
Adding no:label will show everything without a label.