ngxson/llama.cpp

Pull Requests Commits

Merge branch 'master' into xsn/server_tools_shell_stream

ngxson committed 27 days ago

6a0f329d

ngxson committed 27 days ago

44a53642

server: remove loading.html (#25500)

ngxson committed 27 days ago

Verified a935fbff

ngxson committed 27 days ago

888a1c0a

ngxson committed 27 days ago

75aeda4f

add stream for shell tool

ngxson committed 27 days ago

99e7119f

ngxson committed 27 days ago

9431d1d9

ggerganov committed 27 days ago

0badc06a

ggml : use ggml_vqtbl1q_u8 for 32-bit compat (whisper/0)

ggerganov committed 27 days ago

ac17f8ac

server: improve tools, remove apply_diff (#25498)

ngxson committed 27 days ago

Verified c4ae9a88

cli: fix crash on wrong server base url (#25497)

marcoStocchi committed 27 days ago

Verified 1b9691bc

ui: prevent tooltip from flickering open and closed on hover (#25503)

ServeurpersoCom committed 27 days ago

Verified c7af942e

sync : ggml (#25517)

ggerganov committed 27 days ago

Verified 8f114a9b

ui: export full message tree instead of active path only (#25501)

ServeurpersoCom committed 27 days ago

Verified d46786f2

llama : make all KQ masks f16 if FA is used, remove zero attention bias, remove raw_k repeats in DeepSeek V4 (#25370)

fairydreaming committed 27 days ago

Verified 2ed3c1ab

ggml-et: Initial ET backend (#24179)

marty1885 committed 27 days ago

Verified 082b326f

llama-batch: add unit test (#25471)

am17an committed 27 days ago

Verified 961e4b26

opencl: cluster-parallel decode FA for Adreno (#25473)

wanghqc committed 28 days ago

Verified 049326a0

ggml : process data in smaller chunks in CUDA ggml_top_k() and ggml_argsort() to reduce temporary buffers memory usage (#24776)

fairydreaming committed 28 days ago

Verified 07494499

cli: add --output option (#25484)

ngxson committed 28 days ago

Verified 3de7dd4c

hexagon: tiling, tracing and optimizations for unary ops (#25474)

aparmp-quic committed 28 days ago

Verified fb30ba9a

server : move chat-template thinking probe inside the init try/catch (#24093)

palios-taey committed 28 days ago

Verified 82fce65d

ggml : fix conv 2d dw (#25490)

ggerganov committed 28 days ago

Verified 5c3a5868

meta: add hard emphasis on agents not writing descriptions/comments (#25480)

pwilkin committed 28 days ago

Verified c15c5c77

Refactor: Consistently use smart pointers in `test-backend-ops` (#25440)

ORippler committed 28 days ago

Verified f84a5194

Only index by compile times + always multiply/add (#25445)

ORippler committed 28 days ago

Verified 683f0c72

llama-bench : init params.offline (#25476)

angt committed 28 days ago

Verified 259f2e2a

metal : add CONV_2D_DW (depthwise convolution) support (#21565)

Sou-ly committed 28 days ago

Verified 92b187c9

ggml-hip: enable -funsafe-math-optimizations (#24668)

RapidMark committed 28 days ago

Verified ccb0c342

cuda: align snake fusion matcher with the other backends (#25460)

ServeurpersoCom committed 28 days ago

Verified 2021515a

Older