ggerganov/llama.cpp

Pull Requests Commits

Revert "cuda : use CUDA memory pool with async memory allocation/deallocation when available (#3903)"

slaren committed 1 year ago

3ef358ff

Revert "cuda : add ROCM aliases for CUDA pool stuff (#3918)"

slaren committed 1 year ago

6b10aa9f

metal : round up to 16 to fix MTLDebugComputeCommandEncoder assertion (#3938)

psugihara committed 1 year ago

Verified d9b33fe9

ggml-metal: fix yarn rope (#3937)

jxy committed 1 year ago

Verified 5ba37461

ggml-cuda : move row numbers to x grid dim in mmv kernels (#3921)

slaren committed 1 year ago

Verified abb77e73

speculative : change default p_accept to 0.5 + CLI args (#3919)

ggerganov committed 1 year ago

Verified 8f961abd

common : YAYF (yet another YARN fix) (#3925)

ggerganov committed 1 year ago

Verified 05816027

llama : change yarn_ext_factor placeholder to -1 (#3922)

cebtenzzre committed 1 year ago

Verified 3fdbe6b6

cuda : add ROCM aliases for CUDA pool stuff (#3918)

KerfuffleV2 committed 1 year ago

Verified 629f917c

cmake : fix relative path to git submodule index (#3915)

abetlen committed 1 year ago

Verified 51b2fc11

readme : add notice about #3912

ggerganov committed 1 year ago

Verified 224e7d5b

cuda : fix const ptrs warning causing ROCm build issues (#3913)

ggerganov committed 1 year ago

Verified c7743fe1

cuda : use CUDA memory pool with async memory allocation/deallocation when available (#3903)

young-developer committed 1 year ago

Verified d6069051

gguf : print error for GGUFv1 files (#3908)

ggerganov committed 1 year ago

Verified 4ff1046d

cmake : disable LLAMA_NATIVE by default (#3906)

slaren committed 1 year ago

Verified 21958bb3

gguf : remove special-case code for GGUFv1 (#3901)

ggerganov committed 1 year ago

Verified 2756c4fb

llm : prevent from 1-D tensors being GPU split (#3697)

ggerganov committed 1 year ago

Verified 1efae9b7

build : link against build info instead of compiling against it (#3879)

cebtenzzre committed 1 year ago

Verified b12fa0d1

cuda : check if this fixes Pascal card regression (#3882)

ggerganov committed 1 year ago

Verified 4d719a6d

metal : fix build errors and kernel sig after #2268 (#3898)

ggerganov committed 1 year ago

Verified 183b3fac

cuda : fix RoPE after #2268 (#3897)

cebtenzzre committed 1 year ago

Verified 2fffa0d6

llama : fix llama_context_default_params after #2268 (#3893)

cebtenzzre committed 1 year ago

Verified 0eb332a1

ggml-cuda : compute ptrs for cublasGemmBatchedEx in a kernel (#3891)

slaren committed 1 year ago

Verified d02e98cd

llama : implement YaRN RoPE scaling (#2268)

cebtenzzre committed 1 year ago

Verified 898aeca9

llm : fix llm_build_kqv taking unused tensor (benign, #3837)

ggerganov committed 1 year ago

Verified c43c2da8

llm : fix falcon norm after refactoring (#3837)

ggerganov committed 1 year ago

Verified 523e49b1

metal : multi-simd softmax (#3710)

ggerganov committed 1 year ago

Verified e16b9fa4

common : minor (#3715)

ggerganov committed 1 year ago

Verified ff8f9a88

llm : add llm_build_context (#3881)

ggerganov committed 1 year ago

Verified 50337961

common : allow caller to handle help/argument exceptions (#3715)

bandoti committed 1 year ago

Verified 0e40806c

Older