ggerganov/llama.cpp

Pull Requests Commits

alloc : fix build with debug

ggerganov committed 1 year ago

Verified eb594c0f

metal : simplify soft max kernel

ggerganov committed 1 year ago

Verified d9c8fa3b

metal : warp-based reduce for rms_norm

ggerganov committed 1 year ago

Verified c4db5923

metal : warp-based reduction for soft max kernel

ggerganov committed 1 year ago

Verified 55717c98

cuda : fix warp reduction initialization of shared mem

ggerganov committed 1 year ago

Verified 68e02c0d

cuda : increase max block size to 1024

ggerganov committed 1 year ago

Verified 6b86bcff

cuda : do warp-based block reduce

ggerganov committed 1 year ago

Verified 62532c05

ggml : update soft max cpu

ggerganov committed 1 year ago

Verified c7c8dabc

cuda : use 512 threads for soft_max instead of 32

ggerganov committed 1 year ago

Verified ebd062bc

metal : simplify soft_max encoding

ggerganov committed 1 year ago

Verified 580fe206

batched-bench : print threads

ggerganov committed 1 year ago

Verified 390a4459

ggml : implement soft_max_ext (CPU)

ggerganov committed 1 year ago

Verified 6a66f69f

cuda : implement soft_max_ext

ggerganov committed 1 year ago

Verified 88519fbf

metal : implement soft_max_ext

ggerganov committed 1 year ago

Verified e89597c0

examples : add readme files

ggerganov committed 1 year ago

Verified 1f5cd832

readme : add FreeChat (#4248)

psugihara committed 1 year ago

Verified 4fea3420

ggml : restore abort() in GGML_ASSERT (#4242)

cebtenzzre committed 1 year ago

Verified 64e64aa2

ggml : re-enable BLAS for CPU when src0 != F32 + remove redundant full offload checks in llama.cpp (#4240)

ggerganov committed 1 year ago

Verified 8406b092

cmake : fix issue with version info not getting baked into LlamaConfig.cmake (#3970)

bandoti committed 1 year ago

Verified b38a16df

readme : add Amica to UI list (#4230)

kasumi-1 committed 1 year ago

Verified 0dab8cd7

examples : iOS example with swift ui (#4159)

bachittle committed 1 year ago

Verified bb03290c

ggml : fix -Warray-bounds warning with gcc (#4231)

cebtenzzre committed 1 year ago

Verified f3b26981

lookahead : support `-n -1` infinite generation

ggerganov committed 1 year ago

Verified 3e73d31d

readme : update hot topics

ggerganov committed 1 year ago

Verified 9656026b

lookahead : add example for lookahead decoding (#4207)

ggerganov committed 1 year ago

Verified 922754a8

metal : fix yarn (#4220)

jxy committed 1 year ago

Verified 22da0553

scripts : Use mmap in torch load (#4202)

Galunid committed 1 year ago

Verified 1ddb52ec

llama : grammar `reserve` space in `decode_utf8` (#4210)

MarcusDunn committed 1 year ago

Verified f837c3a9

Update docs for yarn_ext_factor <0.0 as unspecified instead of NaN (#4189)

crasm committed 1 year ago

Verified 3014b541

readme : update hot topics

ggerganov committed 1 year ago

Verified 04814e71

Older