ggerganov/llama.cpp

Pull Requests Commits
CUDA: Fixed 7b q3_K_S with mul_mat_vec_q (#2313)
JohannesGaessler committed 2 years ago
Verified b9b7d94f
llama : optimize memory buffers (#2325)
ggerganov committed 2 years ago
Verified b47b8a9c
Perplexity: Compute scores correlated to HellaSwag (#2312)
klosax committed 2 years ago
Verified b5fe67f8
examples : basic VIM plugin
whoreson committed 2 years ago
Verified 24baa54a
ci : fix args
ggerganov committed 2 years ago
Verified dd6c67d3
ci : add 7B CUDA tests (#2319)
ggerganov committed 2 years ago
Verified 5d500e8c
examples : add easy python script to create quantized (k-bit support) GGML models from local HF Transformer models (#2311)
richardr1126 committed 2 years ago
Verified 7d5f1846
Custom RoPE + bettter memory management for CUDA (#2295)
ikawrakow committed 2 years ago
Verified d924522a
Faster Q3_K implementation on Metal (#2307)
ikawrakow committed 2 years ago
Verified 4d76a5f4
ggml : fix the rope fix (513f8619535a64fa9ace808cdcbcf66211535f5c)
ggerganov committed 2 years ago
Verified 0db14fef
examples : fix typo in minigpt4.py (#2298)
eltociear committed 2 years ago
Verified 03e56697
ggml : fix rope args order + assert (#2054)
ggerganov committed 2 years ago
Verified 513f8619
gitignore : fix final newline
ggerganov committed 2 years ago
Verified 3973b25a
llama : remove cfg smooth factor as it is only a reparameterization of the guidance scale (#2280)
Vermeille committed 2 years ago
Verified ab0e26bd
gitignore : changes for Poetry users + chat examples (#2284)
yukiteruamano committed 2 years ago
Verified 73643f5f
make : fix indentation
ggerganov committed 2 years ago
Verified a814d04f
ci : fix MNT realpath usage (#2250)
ggerganov committed 2 years ago
Verified 4c013bb7
make : support customized LLAMA_CUDA_NVCC and LLAMA_CUDA_CCBIN (#2275)
skyan committed 2 years ago
Verified 42c7c2e2
flake : remove intel mkl from flake.nix due to missing files (#2277)
Freed-Wu committed 2 years ago
Verified 78a3d134
llama : make tensor_split ptr instead of array (#2272)
ggerganov committed 2 years ago
Verified ae178ab4
make : add new target for test binaries (#2244)
jpodivin committed 2 years ago
Verified 54e3bc76
MIKU MAYHEM: Upgrading the Default Model for Maximum Fun 🎉 (#2287)
at8u committed 2 years ago
Verified 019fe257
Faster Q2_K on Metal (#2297)
ikawrakow committed 2 years ago
Verified e68c96f7
make : fix embdinput library and server examples building on MSYS2 (#2235)
przemoc committed 2 years ago
Verified 9cf022a1
Faster Q5_K and Q6_K on Metal (#2294)
ikawrakow committed 2 years ago
Verified e782c9e7
Faster Q4_K on Metal (#2290)
ikawrakow committed 2 years ago
Verified 785829df
llama : fix regression from #2000 - could not load no-mmap models
ggerganov committed 2 years ago
fff0e0ea
metal: minor q4 optimization and reduce code size (#2248)
lshzh-ww committed 2 years ago
Verified 417a85a0
llama : extend API to get max devices at runtime (#2253)
SanftMonster committed 2 years ago
Verified 294f4245
flake : update flake.nix (#2270)
Freed-Wu committed 2 years ago
Verified 45a1b07e
Terms Of Use Privacy Policy Refund Policy Impressum