ggml-org/llama.cpp

Pull Requests Commits
Convert vector to f16 for dequantize mul mat vec (#1913)
JohannesGaessler committed 2 years ago
Verified 16b9cd19
Added tokens per second to info prints (#1928)
JohannesGaessler committed 2 years ago
Verified b24c3049
Fixed incorrectly applying RMS norm twice (#1925)
JohannesGaessler committed 2 years ago
Verified 0ede372a
ggml : fix bug in ggml_compute_forward_add_q_f32 (#1918)
l3utterfly committed 2 years ago
Verified 8596af42
readme : update Android build instructions (#1922)
mikeyang01 committed 2 years ago
Verified e1886cf4
llama : prevent usage of k-quants when tensor size is not a multiple of 256 (#1921)
ikawrakow committed 2 years ago
Verified 8ab8ba62
examples : fix examples/metal (#1920)
ikawrakow committed 2 years ago
Verified 90cc59d6
metal : handle buffers larger than device's maxBufferLength (#1826)
ggerganov committed 2 years ago
Verified ce2c7d72
cmake : add CUDA_ARCHITECTURES to new target ggml_static (#1917)
howard0su committed 2 years ago
Verified 57cd6946
make : do not print help for simple example
ggerganov committed 2 years ago
Verified b2416493
minor : warning fixes
ggerganov committed 2 years ago
Verified 4f9c43e3
Only one CUDA stream per device for async compute (#1898)
JohannesGaessler committed 2 years ago
Verified 2c9380dd
llama : fix kv_cache `n` init (close #1903)
ggerganov committed 2 years ago
Verified 051e1b0e
make : update for latest Arch (#1701)
daniandtheweb committed 2 years ago
Verified 86c75718
ggml : fix warnings under MSVC (#1908)
howard0su committed 2 years ago
Verified 3d59ec59
metal : add norm, cpy f16->f16, alibi kernels (#1823)
apage43 committed 2 years ago
Verified 0711a5f6
exposed modules so that they can be invoked by nix run github:ggerganov/llama.cpp#server etc (#1863)
faezs committed 2 years ago
Verified fc45a81b
Server Example Refactor and Improvements (#1570)
digiwombat committed 2 years ago
Verified 794db3e7
hooks : setting up flake8 and pre-commit hooks (#1681)
jpodivin committed 2 years ago
Verified 5ddf7ea1
readme : alternative way to build for Android with CLBlast. (#1828)
gustrd committed 2 years ago
Verified bac19927
Allow cmake to build ggml as a library (#1896)
KerfuffleV2 committed 2 years ago
Verified b4c6f46f
train : get raw text instead of page with html (#1905)
Davidy22 committed 2 years ago
Verified 92f20d99
opencl : support k-quants (#1836)
0cc4m committed 2 years ago
Verified d411968e
examples : add "simple" (#1840)
SuperUserNameMan committed 2 years ago
Verified b41b4cad
cmake : add auto detection of BLAS_INCLUDE_DIRS (#1886)
zenixls2 committed 2 years ago
Verified 13fe9d2d
llama : fix embd when offloading non-repeating layers (#1891)
JohannesGaessler committed 2 years ago
Verified ac3b8869
Fixed possible macro redefinition (#1892)
FrankHB committed 2 years ago
Verified 5b9ccaf1
build : fix and ignore MSVC warnings (#1889)
iboB committed 2 years ago
Verified 9cbf50c0
CUDA : faster k-quant dot kernels (#1862)
ikawrakow committed 2 years ago
Verified 3d011226
gitignore : add several entries specific to Visual Studio (#1888)
iboB committed 2 years ago
Verified 602c7488
Terms Of Use Privacy Policy Refund Policy Impressum