ggml-org/llama.cpp

Pull Requests Commits

ggml : fix unaligned access in AMX code

ggerganov committed 100 days ago

Verified 96f9f391

ci : fix musa docker build (#16306)

yeahdongcn committed 101 days ago

Verified d9e0e7c8

devops: switch to using ubuntu-22.04-s390x image (#16302)

taronaeo committed 101 days ago

Verified 0124ac98

Fixed a few typos in the README of the LLaMA.cpp HTTP Server [no ci] (#16297)

ImadSaddik committed 101 days ago

Verified 2811c652

vulkan: 64-bit im2col (#16135)

jeffbolznv committed 101 days ago

Verified d8359f5f

metal : extend mat-mat multiplication support (#16225)

ggerganov committed 101 days ago

Verified 6a2c6145

metal : fuse non-sequential nodes (#16102)

ggerganov committed 101 days ago

Verified 3b53634f

vulkan: handle mat_mul with A matrix > 4GB (#16176)

jeffbolznv committed 101 days ago

Verified 1384abf8

vulkan: support arbitrary KV dimension in flash attention (#16160)

jeffbolznv committed 101 days ago

Verified e6d65fb0

vulkan : make the vulkan.hpp dynamic dispatcher instance private (#16224)

Acly committed 101 days ago

Verified 8656f5de

Show message actions by default (#16289)

allozaur committed 101 days ago

Verified 4807e8f9

CUDA: mul_mat_id for mmf for bs <= 64 for f16 and bs <= 32 for f32 (#16277)

am17an committed 101 days ago

Verified c0bfc57a

CUDA: refactor and deduplicate vector FA kernels (#16208)

JohannesGaessler committed 101 days ago

Verified 75a3a6c2

vulkan: throw system error instead of SIGABRT during init on older devices (#16156)

DmyMi committed 101 days ago

Verified 0499b29c

server : remove old LLAMA_SERVER_SSL (#16290)

angt committed 101 days ago

Verified 234e2ff8

vulkan: support GET_ROWS for k-quants (#16235)

jeffbolznv committed 102 days ago

Verified 3f81b4e9

build : add LLAMA_OPENSSL option (#16287)

angt committed 102 days ago

Verified ace6a545

model : make minicpm embedding_scale, residual_scale and logit_scale optional with legacy defaults (#16273)

vinkal-chudgar committed 102 days ago

Verified 72b24d96

devops: add s390x & ppc64le CI (#15925)

taronaeo committed 102 days ago

Verified 624207e6

Enhance text file detection logic for file attachments (#16199)

allozaur committed 102 days ago

Verified 807e8c6d

Allow viewing conversations even when llama server is down (#16255)

allozaur committed 102 days ago

Verified 1a189278

webui: switch to hash-based routing (alternative of #16079) (#16157)

isaac-mcfadyen committed 102 days ago

Verified e0539eb6

Always show message actions for mobile UI + improvements for user message sizing (#16076)

allozaur committed 103 days ago

Verified 5d0a40f3

codeowners : add rgerganov as owner of RPC [no ci] (#16279)

rgerganov committed 103 days ago

Verified d12a9836

mtmd : fix uninitialized variable in bicubic_resize (#16275)

AlekseiNikiforovIBM committed 103 days ago

Verified cc1cfa27

metal : report OOM errors (#16274)

ggerganov committed 103 days ago

Verified 54dbc370

common : use cpp-httplib as a cURL alternative for downloads (#16185)

angt committed 103 days ago

Verified b995a107

build : fix build-ios-device (#16257)

angt committed 103 days ago

Verified 4710dd31

ggml-cpu: implement MXFP4 SIMD for s390x (#16193)

taronaeo committed 103 days ago

Verified 9b265118

ci : create git tags for released docker images (#16008)

rgerganov committed 103 days ago

Verified 00217cd4

Older