ggerganov/llama.cpp

Pull Requests Commits

server : simplify prompt state transition branches

ggerganov committed 22 days ago

Verified 4a275125

fix : nullptr task dereference

ggerganov committed 22 days ago

Verified cc5cafec

cont : reduce parent checks

ggerganov committed 22 days ago

Verified aef22e7a

cont : remove redundant function

ggerganov committed 22 days ago

Verified 9ceb268e

cont : improve n_cmpl logic

ggerganov committed 22 days ago

Verified a4854f03

ggerganov committed 22 days ago

Verified f2d988db

Merge branch 'master' into pr/18663

ggerganov committed 22 days ago

Verified 91fd50be

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

ggerganov committed 22 days ago

Verified 53eb9435

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

ggerganov committed 22 days ago

Verified d3435efc

cont : init child samplers + modify child logic

ggerganov committed 22 days ago

Verified 439c3b50

Merge branch 'master' into HEAD

ggerganov committed 22 days ago

Verified 59dda88a

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

ggerganov committed 22 days ago

Verified f5f8812f

common: support remote preset (#18520)

ngxson committed 23 days ago

Verified 8ece3836

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

taronaeo committed 23 days ago

Verified 046d5fd4

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

yomaytk committed 23 days ago

Verified 480160d4

ggml webgpu: initial flashattention implementation (#18610)

reeselevine committed 23 days ago

Verified 15bff84b

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

jeffbolznv committed 23 days ago

Verified 2524c261

vulkan: optimize ssm_scan (#18630)

jeffbolznv committed 23 days ago

Verified cb14b069

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

angt committed 23 days ago

Verified 55abc393

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

ggerganov committed 23 days ago

Verified f2f6c880

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

dororodoroddo committed 23 days ago

Verified 945bf106

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

JohannesGaessler committed 23 days ago

Verified 64848deb

ggml: add env var GGML_OP_OFFLOAD_MIN_BATCH (#18535)

DocShotgun committed 23 days ago

Verified 9a5724de

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

danbev committed 23 days ago

Verified 9c142e3a

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

danbev committed 23 days ago

Verified df7fb921

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

JTischbein committed 23 days ago

Verified 2038101b

opencl: add FILL op support (#18682)

shaofeiqi committed 23 days ago

Verified 568371a7

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

CISC committed 24 days ago

Verified 5b8844ae

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

CISC committed 24 days ago

Verified 7e16fef0

cuda : fix build on cuda 12.8 (#18672)

olliewalsh committed 24 days ago

Verified f5245b5e

Older