ggerganov/llama.cpp

Pull Requests Commits

arch : add description about LLM_TENSOR_INFOS

ggerganov committed 31 days ago

Verified c6bba89e

cuda : fix UMA detection on discrete GPUs. (#17537)

matt23654 committed 31 days ago

Verified 909072ab

ggml-cpu: aarm64: q4_K repack gemm and gemv implementations (dotprod only) (#17494)

Alcpz committed 31 days ago

Verified cd8370b4

devops: Add build-essential to Ubuntu 26.04 image (#17531)

ericcurtin committed 31 days ago

Verified d21a76ac

gguf-py : skip endian-conversion of MXFP4 data (#17523)

AlekseiNikiforovIBM committed 31 days ago

Verified 4fcd87cf

vulkan : move contiguous checks to device_supports_op (#17490)

Acly committed 31 days ago

Verified b78db3bd

vulkan: use a fixed 1KB buffer for the add_rms_fusion opt (#17514)

jeffbolznv committed 31 days ago

Verified 142df17c

server: enable jinja by default, update docs (#17524)

ngxson committed 32 days ago

Verified e509411c

opencl: add sqr, sqrt, mean and ssm_conv (#17476)

lhez committed 32 days ago

Verified 7cba58bb

Fix chunks being too small with small matrix sizes (#17526)

Alcpz committed 32 days ago

Verified 5449367b

clip: (minicpmv) fix resampler kq_scale (#17516)

hNSBQZ committed 32 days ago

Verified 1d594c29

vulkan: allow graph_optimize for prompt processing workloads (#17475)

jeffbolznv committed 32 days ago

Verified eec1e33a

vulkan: Implement top-k (#17418)

jeffbolznv committed 32 days ago

Verified 879d6737

ggml-cpu : add RISC-V Zvfh impl for ggml_vec_mad_f16 (#17448)

xctan committed 32 days ago

Verified 6ab4e50d

cmake : use EXCLUDE_FROM_ALL to avoid patch-boringssl.cmake (#17520)

angt committed 32 days ago

Verified 2336cc47

ggml : fix ARM feature verification (#17519)

angt committed 32 days ago

Verified e6923caa

HIP: Patch failed testcase in WMMA-MMQ kernels for RDNA 4 (#17502)

jiachengjason committed 32 days ago

Verified 3e18dba9

CANN: Add MROPE and IMROPE support (#17401)

hipudding committed 32 days ago

Verified eeb5605d

chore: upgrade cpp-httplib from v0.27.0 to v0.28.0 (#17513)

o7si committed 32 days ago

Verified f3a848a3

vulkan: Implement GGML_OP_CUMSUM (#17479)

jeffbolznv committed 32 days ago

Verified b3b03a7b

ggml : add ggml_top_k (#17365)

ggerganov committed 33 days ago

Verified 583cb834

convert : fix big-endian conversion (#17431)

AlekseiNikiforovIBM committed 33 days ago

Verified 05872ac8

codeowners : remove slaren (#17492)

slaren committed 33 days ago

Verified 55ab25ca

CANN: supports out_prod operator for F32 and F16 (#17406)

TianHao324 committed 33 days ago

Verified 064c90d8

webui: add rehype plugin to restore HTML in Markdown table cells (#17477)

ServeurpersoCom committed 33 days ago

Verified b1846f1c

vulkan: Use fewer rows for scalar FA when HS is not a multiple of 16 (#17455)

jeffbolznv committed 33 days ago

Verified d414db02

llama: introduce support for model-embedded sampling parameters (#17120)

taronaeo committed 33 days ago

Verified 877566d5

vulkan: more FA details in vk_perf_logger (#17443)

jeffbolznv committed 34 days ago

Verified 3d07caa9

llama : skip output reordering for single token batches (#17466)

danbev committed 34 days ago

Verified 134e6940

HIP: WMMA-MMQ kernels for RDNA 4 (#17156)

jiachengjason committed 34 days ago

Verified 0543f928

Older