ggerganov/llama.cpp

Pull Requests Commits

ci : disable FMA on Mac OS

ggerganov committed 1 year ago

Verified f46db27e

ci : print mac os "sysctl -a"

ggerganov committed 1 year ago

Verified 83efc920

Merge branch 'master' into HEAD

ggerganov committed 1 year ago

Verified 8e9af803

ggml : sync latest (new ops, macros, refactoring) (#2106)

ggerganov committed 1 year ago

Verified ed9a54e5

Add an API example using server.cpp similar to OAI. (#2009)

jwj7140 committed 1 year ago

Verified f257fd25

Remove call to ggml_cuda_mul_mat_get_wsize

sw committed 1 year ago

81f28f25

Simple webchat for server (#1998)

tobi committed 1 year ago

Verified 7ee76e45

Allow old Make to build server. (#2098)

SlyEcho committed 1 year ago

Verified acc111ca

Update Makefile: clean simple (#2097)

ZhouYuChen committed 1 year ago

Verified 23c7c6fc

CI: make the brew update temporarily optional. (#2092)

Green-Sky committed 1 year ago

Verified 698efad5

[ggml] fix index for ne03 value in ggml_cl_mul_f32 (#2088)

gotope committed 1 year ago

Verified 14a2cc71

fix server crashes (#2076)

SlyEcho committed 1 year ago

Verified 1cf14cce

Fix crash of test-tokenizer-0 under Debug build (#2064)

howard0su committed 1 year ago

Verified cc45a7fe

[llama] No need to check file version when loading vocab score (#2079)

howard0su committed 1 year ago

Verified 55dbb915

server: add option to output probabilities for completion (#1962)

WangHaoranRobin committed 1 year ago

Verified d7d2e6a0

Generalize quantize_fns for simpler FP16 handling

sw committed 1 year ago

f9c585f0

ggml : fix build with OpenBLAS (close #2066)

ggerganov committed 1 year ago

46088f72

Better CUDA synchronization logic (#2057)

JohannesGaessler committed 1 year ago

Verified 0bc2cdfc

Test-based VRAM scratch size + context adjustment (#2056)

JohannesGaessler committed 1 year ago

Verified befb3a35

cmake : don't force -mcpu=native on aarch64 (#2063)

dsd committed 1 year ago

Verified b2132270

metal : release buffers when freeing metal context (#2062)

apage43 committed 1 year ago

Verified 2f8cd979

convert : add support of baichuan-7b (#2055)

foldl committed 1 year ago

Verified 471aab6e

llama : fix return value of llama_load_session_file_internal (#2022)

ggerganov committed 1 year ago

Verified 463f2f4c

llama : catch llama_load_session_file_internal exceptions (#2022)

randxie committed 1 year ago

Verified cb44dbc7

embd-input : fix returning ptr to temporary

ggerganov committed 1 year ago

Verified 79f634a1

train : fix compile warning

ggerganov committed 1 year ago

Verified 04606a15

ggml : disable GGML_TASK_INIT and GGML_TASK_FINALIZE by default (#1995)

mqy committed 1 year ago

Verified b1ca8f36

Use unsigned for random seed (#2006)

howard0su committed 1 year ago

Verified b8c8dda7

Porting the improved K-Quant CUDA kernels to OpenCL (#1966)

LostRuins committed 1 year ago

Verified 96a712ca

llama : replacing auto &kv with const auto &kv (#2041)

mendax0110 committed 1 year ago

Verified d3494bb8

Older