ggerganov/llama.cpp

Pull Requests Commits

quantize : fix using combined imatrix GGUFs (multiple datasets) (#14973)

EAddario committed 99 days ago

Verified e9192bec

server : add support for `embd_normalize` parameter (#14964)

danbev committed 99 days ago

Verified 41e78c56

HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes (#14949)

IMbackK committed 99 days ago

Verified ad4a7001

ggerganov committed 99 days ago

e32a4ec6

cmake : Fix BLAS link interface (ggml/1316)

dg0yt committed 99 days ago

e228de94

vulkan : fix 32-bit builds (ggml/1313)

dg0yt committed 99 days ago

73a8e5ca

CUDA: skip masked KV slices for all FA kernels (#14924)

JohannesGaessler committed 99 days ago

Verified 92b8810e

tests : update for LLAMA_SET_ROWS=1 (#14961)

ggerganov committed 99 days ago

Verified 00131d6e

graph : fix stack-use-after-return (#14960)

ggerganov committed 99 days ago

Verified 1e15bfd4

embeddings: fix extraction of CLS pooling results (#14927)

iamlemec committed 100 days ago

Verified a118d802

CANN: update ops docs (#14935)

bachelor-dou committed 100 days ago

Verified 61550f82

HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (#14945)

IMbackK committed 100 days ago

Verified aa79524c

HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (#14930)

IMbackK committed 100 days ago

Verified b77d1117

HIP: Ignore unsupported unroll transformation in fattn-vec (#14931)

IMbackK committed 100 days ago

Verified c7aa1364

common : avoid logging partial messages (which can contain broken UTF-8 sequences) (#14937)

kallewoof committed 100 days ago

Verified 1a67fcc3

CANN: Add ggml_set_rows (#14943)

hipudding committed 100 days ago

Verified 204f2cf1

cuda : add softcap fusion (#14907)

CISC committed 100 days ago

Verified 138b288b

server-bench: make seed choice configurable (#14929)

JohannesGaessler committed 100 days ago

Verified bbd0f917

CUDA: add roll (#14919)

am17an committed 100 days ago

Verified 0a5036be

opencl : add ops docs (#14910)

lhez committed 101 days ago

Verified 8ad7b3e6

test-backend-ops : extend test case filtering (#14865)

tlemo committed 101 days ago

Verified bda62193

llama-bench : use local GPUs along with RPC servers (#14917)

rgerganov committed 101 days ago

Verified c556418b

ggml-cpu : deduplicate scalar implementations (#14897)

xctan committed 101 days ago

Verified db16e283

SYCL: Add set_rows support for quantized types (#14883)

qnixsynapse committed 101 days ago

Verified cd1fce6d

mtmd : add support for Voxtral (#14862)

ngxson committed 101 days ago

Verified 00fa15fe

CUDA: fix pointer incrementation in FA (#14916)

JohannesGaessler committed 101 days ago

Verified 946b1f68

model : add support for SmallThinker series (#14898)

wdl339 committed 101 days ago

Verified 6c6e397a

sycl: refactor quantization to q8_1 (#14815)

Alcpz committed 101 days ago

Verified afc0e896

ops : update BLAS (#14914)

ggerganov committed 101 days ago

Verified a5771c9e

ops : update Metal (#14912)

ggerganov committed 102 days ago

Verified c35f9eaf

Newer Older