ggml-org/llama.cpp

Pull Requests Commits

server : chunked prefill support

ggerganov committed 279 days ago

Verified a6648b9d

llama : add 128k yarn context for Qwen (#10698)

robbiemu committed 279 days ago

Verified 62e84d98

server : (refactor) no more json in server_task input (#10691)

ngxson committed 279 days ago

Verified 3573fa8e

ggml : disable iq4_nl interleave size 8 (#10709)

ggerganov committed 279 days ago

Verified d9c3ba2b

server : various fixes (#10704)

ggerganov committed 279 days ago

Verified ce4a7b84

ggml : refactor online repacking (#10446)

Djip007 committed 279 days ago

Verified 19d8762a

server : fix free of spec context and batch (#10651)

ggerganov committed 280 days ago

Verified c2a16c0b

Vulkan: VK_KHR_cooperative_matrix support to speed up prompt processing (#10597)

0cc4m committed 280 days ago

Verified 3df784b3

metal : Extend how Llama.cpp locates metal resources (#10676)

ormandi committed 280 days ago

Verified 86a19349

convert : add support for Roberta embeddings (#10695)

Ssukriti committed 280 days ago

Verified 784a14aa

convert : add custom attention mapping

ggerganov committed 280 days ago

Verified c5ede384

common : bring back --no-warmup to server (#10686)

ngxson committed 280 days ago

Verified f162d45a

server : (refactoring) do not rely on JSON internally (#10643)

ngxson committed 280 days ago

Verified 6c5bc062

fix(server) : not show alert when DONE is received (#10674)

pminev committed 281 days ago

Verified 7736837d

vulkan: Add VK_NV_cooperative_matrix2 support for mul_mat and flash attention (#10206)

jeffbolznv committed 281 days ago

Verified c9c6e01d

llama : add Minerva 7B model support (#10673)

Riccorl committed 281 days ago

Verified 6fe62478

ggerganov committed 281 days ago

Verified 0cd182eb

ggml: add `GGML_SET` Metal kernel + i32 CPU kernel (ggml/1037)

PABannier committed 281 days ago

Verified a8cbab20

ggml : add `GGML_PAD_REFLECT_1D` operation (ggml/1034)

PABannier committed 281 days ago

Verified c2082d93

py : update outdated copy-paste instructions [no ci] (#10667)

danbev committed 282 days ago

Verified d405804b

Update deprecation-warning.cpp (#10619)

aryantandon01 committed 282 days ago

Verified f112d198

server : fix speculative decoding with context shift (#10641)

ggerganov committed 282 days ago

Verified 1da7b765

ggml : add predefined list of CPU backend variants to build (#10626)

slaren committed 282 days ago

Verified 59f4db10

ggml-cpu : fix HWCAP2_I8MM value (#10646)

slaren committed 282 days ago

Verified 28035408

Fix HF repo commit to clone lora test models (#10649)

ltoniazzi committed 283 days ago

Verified 253b7fde

llama: Support MiniCPM-1B (with & w/o longrope) (#10559)

JFLFY2255 committed 283 days ago

Verified 8d0cfd55

vulkan: Implement "fast divide" (mul+shift) for unary ops like copy (#10642)

jeffbolznv committed 283 days ago

Verified 2759916d

SYCL : Move to compile time oneMKL interface backend selection for NVIDIA backend (#10584)

s-Nick committed 283 days ago

Verified 40c6d79f

fix typo of README.md (#10605)

WrRan committed 283 days ago

Verified 98036d56

Avoid using __fp16 on ARM with old nvcc (#10616)

frankier committed 283 days ago

Verified cd2f37b3

Older