ngxson/llama.cpp

Pull Requests Commits

metal : add MoE kernel specialization for ne20=5 (#18667)

dororodoroddo committed 15 days ago

Verified 945bf106

llama-fit-params: free memory target per device (#18679)

JohannesGaessler committed 15 days ago

Verified 64848deb

ggml: add env var GGML_OP_OFFLOAD_MIN_BATCH (#18535)

DocShotgun committed 15 days ago

Verified 9a5724de

model-conversion : add warn about transformers mismatch (#18691)

danbev committed 15 days ago

Verified 9c142e3a

model-conversion : remove -st targets for converted model (#18689)

danbev committed 15 days ago

Verified df7fb921

llama : add `use_direct_io` flag for model loading (#18166)

JTischbein committed 15 days ago

Verified 2038101b

opencl: add FILL op support (#18682)

shaofeiqi committed 15 days ago

Verified 568371a7

scripts : fix repos cloned with .git extension (#18669)

CISC committed 15 days ago

Verified 5b8844ae

convert : more variants of rope_theta config entries (#18668)

CISC committed 15 days ago

Verified 7e16fef0

cuda : fix build on cuda 12.8 (#18672)

olliewalsh committed 15 days ago

Verified f5245b5e

fix(docker): add missing libglvnd libraries to Vulkan image (#18664)

686f6c61 committed 16 days ago

Verified ae9f8df7

tools : remove llama-run (#18661)

angt committed 16 days ago

Verified 56d2fed2

scripts : add pr2wt.sh (#18644)

ggerganov committed 16 days ago

Verified 56426673

convert : clarify sentence-transformers-dense-modules help [no ci] (#18662)

danbev committed 16 days ago

Verified bb77764c

ci : run cann build unconditionally [no ci] (#18659)

CISC committed 16 days ago

Verified 9dfa8ee9

vulkan: reject ops when a tensor is too large to allocate (#18646)

jeffbolznv committed 16 days ago

Verified ca4a8370

vulkan: Warptile tuning for Intel Xe2/Xe3 (#18178)

virajwad committed 16 days ago

Verified 03023296

vulkan: more mul mat optimizations (#18533)

netrunnereve committed 16 days ago

Verified 8c77a04c

examples : add debug utility/example (#18464)

danbev committed 16 days ago

Verified ffba4f29

CANN: Fix rename for get_env (#18652)

hipudding committed 16 days ago

Verified 3333951d

CANN: Rename `get_env` to `get_env_as_lowercase` (#18624)

rauletorresc committed 16 days ago

Verified 193ee38a

Hexagon add support for f16/f32 flash attention, scale, set-rows and improve f16/32 matmul (#18611)

max-krasnyansky committed 16 days ago

Verified 95ea9e08

mtmd: mtmd_audio_streaming_istft (#18645)

tdakhran committed 16 days ago

Verified ccbc84a5

llama-params-fit: fix last devices with low VRAM (#18494)

JohannesGaessler committed 16 days ago

Verified 68b4d516

ggml : optimize cuda ssm_scan using warp-level reduction (#18505)

Aadeshveer committed 16 days ago

Verified 24af22fc

arg: use CSV escape style for multiple-value args (#18643)

ngxson committed 17 days ago

Verified 07fbe19f

vulkan: support buffer_from_host_ptr (#18467)

jeffbolznv committed 17 days ago

Verified ea13cba8

ggml-cuda: refactor cuda graph usage (#18637)

am17an committed 17 days ago

Verified 090b137e

mmq.cu: tune mmq/rocblas switching for RDNA (#18537)

Beinsezii committed 17 days ago

Verified 96892952

server : add thinking content blocks to Anthropic Messages API (#18551)

686f6c61 committed 17 days ago

Verified 3d26a09d

Newer Older