ngxson/llama.cpp

Pull Requests Commits

test_invalid_input_extra_req

ngxson committed 1 year ago

3a81c606

ngxson committed 1 year ago

055aa9e2

ngxson committed 1 year ago

d47360e5

use another model

ngxson committed 1 year ago

5ffc2a02

ngxson committed 1 year ago

ac2ea538

ngxson committed 1 year ago

a4d25724

ngxson committed 1 year ago

6ec3f77a

server : fix format_infill

ngxson committed 1 year ago

b46dc2f2

Vulkan: fix NaN in tanh.comp with AMD proprietary driver on Windows (#10723)

stduhpf committed 1 year ago

Verified 06d70147

llama : use cmake for swift build (#10525)

slaren committed 1 year ago

Verified 43ed389a

vulkan: compile a test shader in cmake to check for coopmat2 support (#10713)

jeffbolznv committed 1 year ago

Verified ecc93d05

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

robbiemu committed 1 year ago

Verified 62e84d98

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

ngxson committed 1 year ago

Verified 3573fa8e

ggml : disable iq4_nl interleave size 8 (#10709)

ggerganov committed 1 year ago

Verified d9c3ba2b

server : various fixes (#10704)

ggerganov committed 1 year ago

Verified ce4a7b84

ggml : refactor online repacking (#10446)

Djip007 committed 1 year ago

Verified 19d8762a

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

ggerganov committed 1 year ago

Verified c2a16c0b

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

0cc4m committed 1 year ago

Verified 3df784b3

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

ormandi committed 1 year ago

Verified 86a19349

convert : add support for Roberta embeddings (#10695)

Ssukriti committed 1 year ago

Verified 784a14aa

convert : add custom attention mapping

ggerganov committed 1 year ago

Verified c5ede384

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

ngxson committed 1 year ago

Verified f162d45a

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

ngxson committed 1 year ago

Verified 6c5bc062

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

pminev committed 1 year ago

Verified 7736837d

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

jeffbolznv committed 1 year ago

Verified c9c6e01d

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

Riccorl committed 1 year ago

Verified 6fe62478

ggerganov committed 1 year ago

Verified 0cd182eb

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

PABannier committed 1 year ago

Verified a8cbab20

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

PABannier committed 1 year ago

Verified c2082d93

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

danbev committed 1 year ago

Verified d405804b

Older