ngxson/llama.cpp

Pull Requests Commits

ngxson committed 2 years ago

de8a4698

use unordered_set everywhere

ngxson committed 2 years ago

24329aac

small change for handle_slots_action

ngxson committed 2 years ago

83249aae

use res_ok everywhere

ngxson committed 2 years ago

588b4bba

ngxson committed 2 years ago

9f56c176

refactor completions handler

ngxson committed 2 years ago

4a5dbd85

server : remove multitask from server_task

ngxson committed 2 years ago

012d8d8c

llama : support RWKV v6 models (#8980)

MollySophia committed 2 years ago

Verified 8f1d81a0

nix: fix CUDA build - replace deprecated autoAddOpenGLRunpathHook

enolan committed 2 years ago

a47667cf

sgemm : improved Q4_0 and Q8_0 performance via 4xN and Mx4 gemm (#8908)

Srihari-mcw committed 2 years ago

Verified ea5d7478

llama : fix typo in xcda_array_view comment [no ci] (#9132)

danbev committed 2 years ago

Verified 49271efb

llama : fix llama_split_mode enum values in main_gpu document (#9057)

kou committed 2 years ago

Verified 0ab30f8d

Correct typo run_llama2.sh > run-llama2.sh (#9149)

MakeDecisionWorth committed 2 years ago

Verified cddae488

llava : the function "clip" should be int (#9237)

tc-mb committed 2 years ago

Verified 7ea8d80d

Threadpool: take 2 (#8672)

fmz committed 2 years ago

Verified 42c76d13

server : fix crash when error handler dumps invalid utf-8 json (#9195)

kaetemi committed 2 years ago

9f7d4bcf

flake.lock: Update (#9162)

ggerganov committed 2 years ago

Verified 1d1ccce6

docker : build images only once (#9225)

slaren committed 2 years ago

Verified 9fe94cca

docker : update CUDA images (#9213)

slaren committed 2 years ago

Verified 66b039a5

vulkan : fix build (#0)

ggerganov committed 2 years ago

20f1789d

ggerganov committed 2 years ago

231cff5f

Fix minicpm example directory (#9111)

xyb committed 2 years ago

Verified 3246fe84

llama : fix qs.n_attention_wv for DeepSeek-V2 (#9156)

compilade committed 2 years ago

Verified 78eb487b

server : add some missing env variables (#9116)

ngxson committed 2 years ago

Verified a77feb5d

llama : fix ChatGLM4 wrong shape (#9194)

RealJosephus committed 2 years ago

Verified 2e59d61c

llama : fix llama3.1 rope_freqs not respecting custom head_dim (#9141)

nyxkrage committed 2 years ago

Verified 75e1dbba

common : Update stb_image.h to latest version (#9161)

arch-btw committed 2 years ago

Verified ad76569f

ggml : do not crash when quantizing q4_x_x with an imatrix (#9192)

slaren committed 2 years ago

Verified 7d787ed9

metal : separate scale and mask from QKT in FA kernel (#9189)

ggerganov committed 2 years ago

Verified 06658ad7

ggml : add SSM Metal kernels (#8546)

ggerganov committed 2 years ago

Verified fc18425b

Older