ggerganov/llama.cpp

Pull Requests Commits

server : adjust unified KV cache tests

ggerganov committed 65 days ago

Verified caff0fd2

server : fix timing of prompt/generation (#18713)

ggerganov committed 65 days ago

Verified 53eb9435

scripts : pr2wt.sh reset to remote head (#18695)

ggerganov committed 65 days ago

Verified d3435efc

server : use different seeds for child completions (#18700)

ggerganov committed 65 days ago

Verified f5f8812f

common: support remote preset (#18520)

ngxson committed 66 days ago

Verified 8ece3836

llama: use host memory if device reports 0 memory (#18587)

taronaeo committed 66 days ago

Verified 046d5fd4

ggml-webgpu: Fix GGML_MEM_ALIGN to 8 for emscripten. (#18628)

yomaytk committed 66 days ago

Verified 480160d4

ggml webgpu: initial flashattention implementation (#18610)

reeselevine committed 66 days ago

Verified 15bff84b

vulkan: fix push constant size for quantize_q8_1 (#18687)

jeffbolznv committed 66 days ago

Verified 2524c261

vulkan: optimize ssm_scan (#18630)

jeffbolznv committed 66 days ago

Verified cb14b069

vendor : update cpp-httplib to 0.30.0 (#18660)

angt committed 66 days ago

Verified 55abc393

scripts : support chaining commands in pr2wt.sh (#18671)

ggerganov committed 66 days ago

Verified f2f6c880

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

dororodoroddo committed 66 days ago

Verified 945bf106

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

JohannesGaessler committed 66 days ago

Verified 64848deb

ggml: add env var GGML_OP_OFFLOAD_MIN_BATCH (#18535)

DocShotgun committed 66 days ago

Verified 9a5724de

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

danbev committed 66 days ago

Verified 9c142e3a

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

danbev committed 66 days ago

Verified df7fb921

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

JTischbein committed 66 days ago

Verified 2038101b

opencl: add FILL op support (#18682)

shaofeiqi committed 66 days ago

Verified 568371a7

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

CISC committed 67 days ago

Verified 5b8844ae

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

CISC committed 67 days ago

Verified 7e16fef0

cuda : fix build on cuda 12.8 (#18672)

olliewalsh committed 67 days ago

Verified f5245b5e

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

686f6c61 committed 67 days ago

Verified ae9f8df7

tools : remove llama-run (#18661)

angt committed 67 days ago

Verified 56d2fed2

scripts : add pr2wt.sh (#18644)

ggerganov committed 67 days ago

Verified 56426673

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

danbev committed 67 days ago

Verified bb77764c

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

CISC committed 67 days ago

Verified 9dfa8ee9

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

jeffbolznv committed 67 days ago

Verified ca4a8370

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

virajwad committed 67 days ago

Verified 03023296

vulkan: more mul mat optimizations (#18533)

netrunnereve committed 67 days ago

Verified 8c77a04c

Older