root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-11-11 21:39:52 +00:00
Code Issues Actions 12 Packages Projects Releases Wiki Activity

convert-hf : match model part name prefix and suffix (#7687)

Browse Source 
In #7075, to fix the conversion of (some) models using model-00001-of-00001.safetensors instead of model.safetensors for a single model part we simply used the same logic as the part count to get the part names. 

But this doesn't always work correctly, like when unusual additional model files like consolidated.safetensors in https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3 are present.

This commit matching both the prefix and the suffix of the model part names should fix this problem without breaking any previously-supported upstream models. But according to report by @teleprint-me there is still some
persistent problem, but shall do in the meantime.
This commit is contained in:
compilade 2024-06-08 22:47:25 -04:00 committed by GitHub
parent ed9f252118
commit 5795b94182
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
convert-hf-to-gguf.py
View File

@ -73,10 +73,10 @@ class Model:
self.endianess = gguf.GGUFEndian.BIG if is_big_endian else gguf.GGUFEndian.LITTLE self.endianess = gguf.GGUFEndian.BIG if is_big_endian else gguf.GGUFEndian.LITTLE
self.use_temp_file = use_temp_file self.use_temp_file = use_temp_file
self.lazy = not eager self.lazy = not eager
self.part_names = Model.get_model_part_names(self.dir_model, ".safetensors") self.part_names = Model.get_model_part_names(self.dir_model, "model", ".safetensors")
self.is_safetensors = len(self.part_names) > 0 self.is_safetensors = len(self.part_names) > 0
if not self.is_safetensors: if not self.is_safetensors:
self.part_names = Model.get_model_part_names(self.dir_model, ".bin") self.part_names = Model.get_model_part_names(self.dir_model, "pytorch_model", ".bin")
self.hparams = Model.load_hparams(self.dir_model) self.hparams = Model.load_hparams(self.dir_model)
self.block_count = self.find_hparam(["n_layers", "num_hidden_layers", "n_layer"]) self.block_count = self.find_hparam(["n_layers", "num_hidden_layers", "n_layer"])
self.tensor_map = gguf.get_tensor_name_map(self.model_arch, self.block_count) self.tensor_map = gguf.get_tensor_name_map(self.model_arch, self.block_count)
@ -335,10 +335,10 @@ class Model:
self.gguf_writer.close() self.gguf_writer.close()
@staticmethod @staticmethod
def get_model_part_names(dir_model: Path, suffix: str) -> list[str]: def get_model_part_names(dir_model: Path, prefix: str, suffix: str) -> list[str]:
part_names: list[str] = [] part_names: list[str] = []
for filename in os.listdir(dir_model): for filename in os.listdir(dir_model):
if filename.endswith(suffix): if filename.startswith(prefix) and filename.endswith(suffix):
part_names.append(filename) part_names.append(filename)
part_names.sort() part_names.sort()