ggerganov/llama.cpp

Pull Requests Commits

CUDA: add softmax broadcast (#14475)

am17an committed 165 days ago

55a1c5a5

CUDA: broadcasting for FlashAttention mask (#14500)

JohannesGaessler committed 165 days ago

12a81af4

vulkan: support softmax/FA batch and broadcast (#14449)

jeffbolznv committed 165 days ago

8875523e

ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (#14435)

ggerganov committed 165 days ago

ec68e84c

opencl : fix possible buffer overflow in dump_tensor (#14490)

jeffzhou2000 committed 165 days ago

Verified 307e79d3

simple-chat : fix context-exceeded condition (#14494)

ggerganov committed 165 days ago

Verified d7f5f4e5

opencl : skip empty nodes on cgraph compute (#14491)

EZForever committed 165 days ago

Verified c8a4e470

opencl : update upscale to support align corners (#14488)

lhez committed 165 days ago

Verified 603e43dc

ci : add OpenCL to labeler workflow (#14496)

CISC committed 165 days ago

Verified 611ba4b2

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

EZForever committed 165 days ago

Verified 85841e12

ggml : Callback before abort (#14481)

ScaledLizard committed 165 days ago

Verified 68b3cd65

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

ggerganov committed 166 days ago

Verified de569441

Add Vulkan images to docker.md (#14472)

xek committed 166 days ago

Verified 1b2aaf28

CANN: update aclnnGroupedMatmulV2 to aclnnGroupedMatmulV3 (#14411)

noemotiovon committed 166 days ago

Verified 343b6e94

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

jeffbolznv committed 166 days ago

Verified 6a746cf9

add GELU_ERF (#14455)

CISC committed 166 days ago

Verified eff5e454

ggml : remove trailing whitespace (#0)

ggerganov committed 166 days ago

a6a47958

ggerganov committed 166 days ago

f61c05d4

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

Acly committed 166 days ago

431b2c24

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

danbev committed 166 days ago

497be7c0

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

lhez committed 166 days ago

Verified 79b33b23

Add Conv2d for CPU (#14388)

am17an committed 166 days ago

Verified 0a5a3b5c

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

ggerganov committed 167 days ago

Verified 745f11fe

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

ggerganov committed 167 days ago

Verified 5dd942de

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

Rbiessy committed 167 days ago

Verified a7417f55

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

slaren committed 167 days ago

Verified eb3fa291

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

xiaobing318 committed 167 days ago

Verified c839a2da

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

vedranmiletic committed 167 days ago

Verified e9b6350e

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

matteoserva committed 167 days ago

Verified caf5681f

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

rntk committed 167 days ago

Verified 83790b0e

Newer Older