alihan/llama-cpp-python
1
0
Fork 0
mirror of https://github.com/abetlen/llama-cpp-python.git synced 2023-09-07 17:34:22 +03:00
Code Issues Packages Projects Releases Wiki Activity

Update llama.cpp

Browse Source
This commit is contained in:
Andrei Betlen
2023-04-21 17:40:27 -04:00
parent ba3959eafd
commit 1eb130a6b2
2 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
llama_cpp/llama_cpp.py
View File

@@ -117,6 +117,8 @@ LLAMA_FTYPE_MOSTLY_Q4_1 = ctypes.c_int(3) # except 1d tensors
LLAMA_FTYPE_MOSTLY_Q4_1_SOME_F16 = ctypes.c_int( LLAMA_FTYPE_MOSTLY_Q4_1_SOME_F16 = ctypes.c_int(
4 4
) # tok_embeddings.weight and output.weight are F16 ) # tok_embeddings.weight and output.weight are F16
LLAMA_FTYPE_MOSTLY_Q4_2 = ctypes.c_int(5) # except 1d tensors
LLAMA_FTYPE_MOSTYL_Q4_3 = ctypes.c_int(6) # except 1d tensors
# Functions # Functions
@@ -169,11 +171,12 @@ _lib.llama_free.restype = None
# TODO: not great API - very likely to change # TODO: not great API - very likely to change
# Returns 0 on success # Returns 0 on success
def llama_model_quantize(fname_inp: bytes, fname_out: bytes, itype: c_int) -> c_int: # nthread - how many threads to use. If <=0, will use std::thread::hardware_concurrency(), else the number given
return _lib.llama_model_quantize(fname_inp, fname_out, itype) def llama_model_quantize(fname_inp: bytes, fname_out: bytes, ftype: c_int, nthread: c_int) -> c_int:
return _lib.llama_model_quantize(fname_inp, fname_out, ftype, nthread)
_lib.llama_model_quantize.argtypes = [c_char_p, c_char_p, c_int] _lib.llama_model_quantize.argtypes = [c_char_p, c_char_p, c_int, c_int]
_lib.llama_model_quantize.restype = c_int _lib.llama_model_quantize.restype = c_int

2
vendor/llama.cpp vendored

Submodule vendor/llama.cpp updated: c8c2c52482...50cb666b8a