ggerganov/llama.cpp

Pull Requests Commits

cuda : graceful fallback for Mamba-1 models with weird embd size

compilade committed 64 days ago

71bef665

Merge branch 'master' into compilade/mamba2

compilade committed 65 days ago

73de1fd1

github : add OpenCL backend to issue templates (#14492)

EZForever committed 65 days ago

Verified 85841e12

ggml : Callback before abort (#14481)

ScaledLizard committed 65 days ago

Verified 68b3cd65

ci : disable fast-math for Metal GHA CI (#14478)

ggerganov committed 65 days ago

Verified de569441

Add Vulkan images to docker.md (#14472)

xek committed 65 days ago

Verified 1b2aaf28

CANN: update aclnnGroupedMatmulV2 to aclnnGroupedMatmulV3 (#14411)

noemotiovon committed 65 days ago

Verified 343b6e94

vulkan: Split large mul_mat_id to fit in shared memory (#14451)

jeffbolznv committed 65 days ago

Verified 6a746cf9

add GELU_ERF (#14455)

CISC committed 65 days ago

Verified eff5e454

ggml : remove trailing whitespace (#0)

ggerganov committed 65 days ago

a6a47958

ggerganov committed 65 days ago

f61c05d4

ggml-cpu : "align corners" for bilinear upscale/downscale (ggml/1285)

Acly committed 65 days ago

431b2c24

ggml-quants : rename best_mad to best_error (ggml/1283)

danbev committed 65 days ago

497be7c0

opencl : add GEGLU, REGLU, SWIGLU (#14456)

lhez committed 65 days ago

Verified 79b33b23

Add Conv2d for CPU (#14388)

am17an committed 66 days ago

Verified 0a5a3b5c

memory : correctly handle failure in apply() (#14438)

ggerganov committed 66 days ago

Verified 745f11fe

metal : disable fast-math for some cpy kernels (#14460)

ggerganov committed 66 days ago

Verified 5dd942de

ggml-cpu: sycl: Re-enable exp f16 (#14462)

Rbiessy committed 66 days ago

Verified a7417f55

test-backend-ops : disable llama test (#14461)

slaren committed 66 days ago

Verified eb3fa291

cmake : Remove redundant include path in CMakeLists.txt (#14452)

xiaobing318 committed 66 days ago

Verified c839a2da

scripts : make the shell scripts cross-platform (#14341)

vedranmiletic committed 66 days ago

Verified e9b6350e

server : support jinja extra template kwargs (Qwen3 enable_thinking feature), from command line and from client (#13196)

matteoserva committed 67 days ago

Verified caf5681f

server : fix appearance of the chats list context menu for Safari (#14322)

rntk committed 67 days ago

Verified 83790b0e

SYCL: disable faulty fp16 exp kernel (#14395)

qnixsynapse committed 67 days ago

Verified f47c1d71

ggml : fix unmerged GGML_FPxx_TO_FPxx refactoring (#14443)

CISC committed 67 days ago

Verified a5d1fb62

ggml : implement REGLU/GEGLU/SWIGLU ops (#14158)

CISC committed 67 days ago

Verified a0535ffa

vulkan: Add fusion support for RMS_NORM+MUL (#14366)

jeffbolznv committed 67 days ago

Verified bd9c981d

CUDA: add bf16 and f32 support to cublas_mul_mat_batched (#14361)

am17an committed 68 days ago

Verified 27208bf6

vulkan: handle noncontig in the final case of ggml_vk_get_cpy_pipeline (#14378)

jeffbolznv committed 68 days ago

Verified 63a7bb3c

vulkan: lock accesses of pinned_memory vector (#14333)

jeffbolznv committed 68 days ago

Verified 00d5282c

Older