root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 03:44:35 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

gguf.py : no need to convert tensors twice

Browse Source
This commit is contained in:
klosax 2023-08-12 21:45:45 +02:00 committed by GitHub
parent 60d540831b
commit 5d81a715d4
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
gguf.py
View File

@ -179,20 +179,20 @@ class GGUFWriter:
def ggml_pad(x: int, n: int) -> int:
return ((x + n - 1) // n) * n
def add_tensor_info(self, name: str, tensor: np.ndarray):
def add_tensor_info(self, name: str, tensor_shape: np.ndarray, tensor_dtype: np.dtype, tensor_nbytes: int):
encoded_name = name.encode("utf8")
self.ti_data += struct.pack("<I", len(encoded_name))
self.ti_data += encoded_name
n_dims = len(tensor.shape)
n_dims = len(tensor_shape)
self.ti_data += struct.pack("<I", n_dims)
for i in range(n_dims):
self.ti_data += struct.pack("<I", tensor.shape[n_dims - 1 - i])
self.ti_data += struct.pack("<I", tensor_shape[n_dims - 1 - i])
assert tensor.dtype in (np.float32, np.float16), "Only F32 and F16 tensors are supported for now"
dtype = GGMLQuantizationType.F32 if tensor.dtype == np.float32 else GGMLQuantizationType.F16
assert tensor_dtype in (np.float32, np.float16), "Only F32 and F16 tensors are supported for now"
dtype = GGMLQuantizationType.F32 if tensor_dtype == np.float32 else GGMLQuantizationType.F16
self.ti_data += struct.pack("<I", dtype)
self.ti_data += struct.pack("<Q", self.offset_tensor)
self.offset_tensor += GGUFWriter.ggml_pad(tensor.nbytes, self.data_alignment)
self.offset_tensor += GGUFWriter.ggml_pad(tensor_nbytes, self.data_alignment)
self.ti_data_count += 1
def write_tensor_to_file(self, tensor: np.ndarray):