mirror of
https://github.com/abetlen/llama-cpp-python.git
synced 2023-09-07 17:34:22 +03:00
Update llama.cpp
This commit is contained in:
@@ -114,6 +114,7 @@ LLAMA_FTYPE_ALL_F32 = ctypes.c_int(0)
|
||||
LLAMA_FTYPE_MOSTLY_F16 = ctypes.c_int(1) # except 1d tensors
|
||||
LLAMA_FTYPE_MOSTLY_Q4_0 = ctypes.c_int(2) # except 1d tensors
|
||||
LLAMA_FTYPE_MOSTLY_Q4_1 = ctypes.c_int(3) # except 1d tensors
|
||||
LLAMA_FTYPE_MOSTLY_Q4_1_SOME_F16 = ctypes.c_int(4) # tok_embeddings.weight and output.weight are F16
|
||||
|
||||
# Functions
|
||||
|
||||
|
||||
2
vendor/llama.cpp
vendored
2
vendor/llama.cpp
vendored
Submodule vendor/llama.cpp updated: 8b679987cd...e7f6997f89
Reference in New Issue
Block a user