ngxson/llama.cpp

Pull Requests Commits

ngxson committed 23 days ago

54b0e6b4

ngxson committed 23 days ago

b80eaf76

metal kernel ok

ngxson committed 23 days ago

bcc50b09

add to test-backend-ops

ngxson committed 23 days ago

adf9529c

mrope now requires 4D tensor

ngxson committed 23 days ago

6814ed7f

chore: update webui build output

ngxson committed 23 days ago

8da55064

ngxson committed 23 days ago

957c0ab4

Merge branch 'master' into xsn/rope_v2

ngxson committed 24 days ago

e4782fb5

ggml-cuda: fix regex for arch list (#18371)

am17an committed 24 days ago

Verified 85c40c9b

cuda: optimize cumsum cub path (#18362)

am17an committed 24 days ago

Verified 83b3b1c2

ggml-cuda: fix blackwell native builds (#18361)

am17an committed 24 days ago

Verified b0fb0f0a

CANN: Add support for CONV_TRANSPOSE_1D when kernel size > 255 (#17934)

Intellouis committed 24 days ago

Verified e68c19b0

ggml : optimize cuda cumsum fallback kernel (#18343)

Aadeshveer committed 24 days ago

Verified c54bba86

server: (router) add stop-timeout option (#18350)

ngxson committed 25 days ago

Verified f5acfb2f

model: support MiMo-V2-Flash (#18328)

ngxson committed 25 days ago

Verified 4cbafad4

fit-params : fix race condition in fit-params output (#18276)

Aadeshveer committed 25 days ago

Verified c1842842

CUDA: experimental native mxfp4 support for blackwell (#17906)

am17an committed 25 days ago

Verified c8a2417d

model : support for LlamaBidirectionalModel architecture (#18220)

sfallah committed 25 days ago

Verified 54132f1b

vulkan: fix command buffer corruption in ggml_backend_vk_event_wait (#18302)

jeffbolznv committed 25 days ago

Verified 2a9ea202

CANN : refactor ACL graph cache (#17752)

wangweixuan committed 25 days ago

Verified ce7a6dc0

docs: Fix typos in SYCL documentation (#18269)

yoka committed 25 days ago

Verified 1ce0126b

vulkan: use fewer FA rows for small cache runs (#18280)

0cc4m committed 25 days ago

Verified 7f459c98

CANN: Uses yarn_ramp cache in ROPE (#17725)

TianHao324 committed 25 days ago

Verified cf2ffc02

common: add `LLAMA_ARG_OVERRIDE_TENSOR` env var for `-ot` arg (#18267)

ddh0 committed 25 days ago

Verified 10355dc7

server: return_progress to also report 0% processing state (#18305)

ngxson committed 26 days ago

Verified 5ee4e43f

webui: apply webui_settings on first load (#18223)

ServeurpersoCom committed 26 days ago

Verified 5b6c9bc0

server: fix crash with model not having BOS/EOS (#18321)

ngxson committed 26 days ago

Verified 849d0211

model-conversion : add device option to run-org-model.py (#18318)

danbev committed 26 days ago

Verified 8e3ead6e

rpc : add check for rpc buffer type (#18242)

struct committed 26 days ago

Verified 12ee1763

ggml-hexagon: create generalized functions for cpu side op (#17500)

chraac committed 26 days ago

Verified ed759777

Older