Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-01-07 11:23:56 +01:00
Code Issues Packages Projects Releases Wiki Activity

fix conflicts

Browse Source
This commit is contained in:
M. Yusuf Sarıgöz 2023-08-13 16:55:42 +03:00
parent 6beebf3fd9
commit 24f48833ab
1 changed files with 38 additions and 44 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
convert-llama-h5-to-gguf.py
View File

@ -23,7 +23,6 @@ def permute(weights: NDArray, n_head: int) -> NDArray:
.swapaxes(1, 2)
.reshape(weights.shape))
def count_model_parts(dir_model: str) -> int:
num_parts = 0
for filename in os.listdir(dir_model):
@ -34,7 +33,6 @@ def count_model_parts(dir_model: str) -> int:
print("gguf: found " + str(num_parts) + " model parts")
return num_parts
if len(sys.argv) < 3:
print("Usage: convert-h5-to-ggml.py dir-model ftype\n")
print(" ftype == 0 -> float32")
@ -261,16 +259,12 @@ for part_name in part_names:
for name in model_part.keys():
data = model_part[name]
<< << << < HEAD
n_dims = len(data.shape)
data_dtype = data.dtype
== == == =
old_dtype = data.dtype
# we don't need these
if name.endswith(".rotary_emb.inv_freq"):
continue
>>>>>> > 17800cd80fec468411481dc34a51d42a936442f1
# convert any unsupported data types to float32
if data.dtype != torch.float16 and data.dtype != torch.float32: