ggml-org/llama.cpp

Pull Requests Commits

qwen3.5: merge tensors on convert also

ORippler committed 1 hour ago

9917a1b5

ci : remove dead snakes (#32084)

CISC committed 4 hours ago

Verified 956591f4

llama : remove non-fused GDN and ubatch-dependent graph nodes (#32078)

ggerganov committed 5 hours ago

Verified ccd360f0

devops : add missing MKL runtime DLLs to the Windows SYCL release package (#32005)

Asahi-Prv committed 5 hours ago

Verified b326dd67

vendor: update cpp-httplib to 0.66.0 (#32049)

cabelo committed 6 hours ago

Verified b8abe41a

backend : re-instantiate CUDA graphs when the cluster dims of a kernel change (#31284)

tdakhran committed 6 hours ago

Verified 2e325dcd

test-fp8.cpp: use only the public ggml API (#32083)

ORippler committed 7 hours ago

Verified 1648d641

ROCm/HIP: fixed BLAS workspace instead of rocBLAS allocating one per call (#31838)

fox5611 committed 7 hours ago

Verified 43243363

quantize: fix imatrix GGUF filepath metadata leak (#32000)

JamieBordoni committed 7 hours ago

Verified 71a6a114

cpu: accumulate flash attention V in F32 (#29810)

ServeurpersoCom committed 8 hours ago

Verified 7f24aacd

webgpu: add PAD_REFLECT_1D support (#31877)

LeON-Nie-code committed 8 hours ago

Verified af6b5243

vulkan: allow dequantized KV optimization with CM2 (#31526)

jeffbolznv committed 8 hours ago

Verified d2b3eb9b

vulkan: allocate the device fault vendor binary buffer (#32008)

figkcoci-debug committed 8 hours ago

Verified eeb21047

tests : use only the public ggml API (#32077)

ORippler committed 8 hours ago

Verified 03d6c824

ui: open the virtual keyboard in the chat editor on Android (#32060)

ServeurpersoCom committed 9 hours ago

Verified adde94c1

cuda : use MMVF for BF16 matrices of few rows at small batch sizes (#31959)

am17an committed 10 hours ago

Verified 459ab1cd

cuda : add sm70 FlashAttention config case for DKQ=256, DV=256, ncols=64 (#27997)

mistrjirka committed 10 hours ago

Verified 77e2db8c

cuda : fattn mma head sizes 40 72 (#30285)

sihanyu03 committed 10 hours ago

Verified 9d7b2d73

CUDA: Add OCP FP8 support (#28413)

ORippler committed 10 hours ago

Verified 9e9619d0

sycl : fixed sycl CLAMP to actually support strided view (#31784)

mctylr-gh committed 13 hours ago

Verified f3b407ca

chat : fix Muse Spark ignoring response_format json_schema (#32017)

rmz-24 committed 20 hours ago

Verified aa75ff32

ui: keep a hidden sidebar from intercepting clicks again (#32035)

allozaur committed 21 hours ago

Verified b5cb8f6f

ui: update the README banner to the new llama.cpp cover and remove the legacy brand assets (#31789)

allozaur committed 21 hours ago

Verified 85822a7a

ui: show audio in Discover for models whose mmproj has an audio encoder (#31837)

allozaur committed 21 hours ago

Verified 2b66bfbf

ci : bump ty to 0.0.100 (#32030)

CISC committed 22 hours ago

Verified 582eab71

mtmd: fix string format, rename to mtmd_format (#32040)

ngxson committed 23 hours ago

Verified 7c2653ac

ui: move Models, MCP and Settings to a sidebar footer (#32022)

allozaur committed 1 day ago

Verified 7592b61d

cuda : remove unused ids and type_src0 parameters (#32018)

danbev committed 1 day ago

Verified cd4ed902

llama : Add FP8 KV-cache support (#28392)

ORippler committed 1 day ago

Verified 202c1a22

Merge pull request #32015 from ggml-org/gg/fix-sampling-output-aliasing

ggerganov committed 1 day ago

Verified e24d0b4c

Older