ggml-org/llama.cpp

Pull Requests Commits

engine: separate engine and server

ngxson committed 22 hours ago

5e599821

model : Granite Switch Decision (#30676)

gabe-l-hart committed 22 hours ago

Verified 23f7aabc

common: honor HTTP_PROXY/HTTPS_PROXY env vars in http client (#21752)

texasich committed 1 day ago

Verified 7182ef71

convert : respect Mistral format for DeepseekV2Model (#30669)

aiqubits committed 1 day ago

Verified 2a667032

Add more benchmarks to llama-eval (#26174)

pwilkin committed 1 day ago

Verified f3363593

hexagon: correct HMX FA sinks (#30581)

jhen0409 committed 1 day ago

Verified ceefd16d

model : MTP support for Mellum2.1 (#30612)

Xarbirus committed 1 day ago

Verified 9a16f1db

convert : speed up model endian conversion and add more datatype support (#30647)

taronaeo committed 1 day ago

Verified b84ea5cf

[SYCL]: fix memory errors in mul_mat, split buffer, host pool (#29889)

mctylr-gh committed 1 day ago

Verified a30fb492

gguf-py : add numpy Q1_0 and Q2_0 de/quantization support (#30630)

SomeNobody21112 committed 1 day ago

Verified 495c5805

CUDA: HIP: rms_norm: add variants that load float4 at a time and store x in registers (#29720)

IMbackK committed 1 day ago

Verified 623c1639

common : read ngram cache parts by reference instead of copying (#30061)

jadidbourbaki committed 1 day ago

Verified 13b80489

convert : support text-only checkpoints in pplx-decider (#30575)

final-bench committed 1 day ago

Verified 0f437626

cpu : reduce AVX2 tiled accumulators together (#30296)

trivalik committed 1 day ago

Verified 3dbcb7fc

ggml-cpu: refactor s309x repack helper functions (#30588)

taronaeo committed 1 day ago

Verified c8047545

sycl : reduce the build threads to half (#30236)

arthw committed 1 day ago

Verified c0083e30

sycl: wide stores for the q4_K/q5_K weight dequant (#29696)

Titaniumtown committed 1 day ago

Verified 13de11e7

tests : catch vanished rows in the fusion baseline (#30618)

ggerganov committed 1 day ago

Verified 88aaf325

vulkan: fix gpu-visible host memory issues (#30468)

0cc4m committed 1 day ago

Verified a8d9a7e1

Fix failing RocM tests (#30611)

mailkv23 committed 1 day ago

Verified f12aaa34

graph : add precision handling for MELLUM architecture (#30584)

Xarbirus committed 1 day ago

Verified 5e03a1ff

hexagon: tensor repacking rework and support for tiled and flat quantized and fp weights (#30635)

max-krasnyansky committed 1 day ago

Verified b797b9ca

llama : share the k-pool input views between layers (#30613)

am17an committed 1 day ago

Verified 6afa9a31

Limit max CUDA pinned host buffer size to 4 GB (#30449)

gaugarg-nv committed 1 day ago

Verified b6334e37

sycl: support reordered Q8_0 MoE weights (#29337)

newjordan committed 1 day ago

Verified e707790c

[SYCL] Enable ssl in build script and CI (#30456)

arthw committed 1 day ago

Verified 43d57fb8

mtmd: add mtmd_audio_stream + support voxtral-realtime (#30629)

ngxson committed 1 day ago

Verified 2851b3bc

convert : fix Granite Switch block_count for MultiSwitch checkpoints (#30621)

planetf1 committed 1 day ago

Verified 59d5fad7

convert : fix regression in mapping of Granite speech tensors (#30628)

mrutkows committed 1 day ago

Verified e287bd87

mtmd: lightonocr does not require longest edge to always be 1540, native resolution is the default (#30600)

staghado committed 1 day ago

Verified 2f729cfa

Older