ggml-org/llama.cpp

Pull Requests Commits

ngxson committed 4 days ago

b8addd2f

common : reduce recursion limit for peg parser

aldehir committed 4 days ago

Verified 510bcb37

grammar : move GRAMMAR_RECURSIVE_LIMIT into the .cpp

ServeurpersoCom committed 4 days ago

9e98dbeb

common : add depth_guard for schema walks

aldehir committed 4 days ago

Verified 1425b1bd

nit : shorten the comment of llama_grammar_reject_step

ServeurpersoCom committed 4 days ago

f7c4801c

grammar : track rejected candidates as indices into the input

ServeurpersoCom committed 4 days ago

47112ac5

grammar : reject candidates in a loop, without a depth limit

ServeurpersoCom committed 4 days ago

8fa4bd1c

common : tidy up peg parser changes

aldehir committed 4 days ago

Verified 5f2f3051

catch more recursive functions

ngxson committed 4 days ago

34f57790

common, grammar: add recursive depth limit, make it a requirement

ngxson committed 4 days ago

91a38ed6

llama: fix exhausting CPU and MEM caused by dry sequence breaker (#31035)

trunk-trick committed 4 days ago

Verified 63412d87

llama-quant : fix 4D tensor quantization using ggml_nrows() and name-based expert detection (#30820)

anthony-wss committed 4 days ago

Verified 5c965fd8

grammar: limit recursion depth to prevent stack overflow (#29495) (#29755)

Hrushikesh-ramilla committed 4 days ago

Verified f1168ac4

llama : respect no_alloc in all memory caches (#31087)

ggerganov committed 4 days ago

Verified fa3645fe

tools : reject truncated GGUF tensor data in merge (#31044)

shreya-11733 committed 4 days ago

Verified 7790a923

CUDA: fuse FFN gate/up matmuls and GLU in MMQ with interleaved gate/up rows (#29948)

anujj committed 4 days ago

Verified c49edf09

ggml-cuda: fuse qwen MRoPE + set-rows (#31025)

am17an committed 4 days ago

Verified a55c16e4

vulkan: Add a larger tile size for conv3d (coopmat2 only) (#29920)

jeffbolznv committed 4 days ago

Verified e321bdad

vulkan : one warp tile per subgroup in the small matmul tiles (#30106)

ianloic committed 4 days ago

Verified 459a1ba5

hexagon: generic dims collapse/reshape for unary, binary and matmul ops (#31071)

max-krasnyansky committed 4 days ago

Verified f0fb5984

opencl: add topk and lightning indexer (#31026)

lhez committed 4 days ago

Verified 9a4580b6

opencl: add initial q2_k support (#30999)

lhez committed 4 days ago

Verified b083c384

ci : update CPU job assignments (#31136)

ggerganov committed 4 days ago

Verified 344be0a6

chat : reject JSON nested more than 128 levels before walking it (#30242)

ykhrustalev committed 4 days ago

Verified b3fa6995

model: DSV4.1 flash (#31086)

am17an committed 5 days ago

Verified 74bca395

jinja : fix select and selectattr argument handling (#31059)

JoeyPatricio committed 5 days ago

Verified d327bf1f

hexagon: KV-head team split for multicore HMX flash attention (#31055)

ebateni committed 5 days ago

Verified 78a8ae14

server: refactor server entrypoint (#31113)

ngxson committed 5 days ago

Verified f3269ef9

metal : remove mul_mm_id amax rescaling (#30924)

ggerganov committed 5 days ago

Verified 3f03b177

rpc : add bounds check in rdma_recv (#30399)

sohaibharraoui committed 5 days ago

Verified 9a0fde99

Older