vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

convert-hf : minor changes for consistency

Browse source
This commit is contained in:
Francis Couture-Harpin 2024-05-05 16:49:16 -04:00
parent bc78bf4cdb
commit 62303e7f77
1 changed files with 5 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

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

@ -2079,6 +2079,8 @@ class GemmaModel(Model):
self.gguf_writer.add_file_type(self.ftype)
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
del bid # unused
# lm_head is not used in llama.cpp, while autoawq will include this tensor in model
# To prevent errors, skip loading lm_head.weight.
if name == "lm_head.weight":
@ -2089,7 +2091,7 @@ class GemmaModel(Model):
if name.endswith("norm.weight"):
data_torch = data_torch + 1
return super().modify_tensors(data_torch, name, bid)
return [(self.map_tensor_name(name), data_torch)]
@Model.register("Starcoder2ForCausalLM")
@ -2277,7 +2279,7 @@ class LazyTorchTensor:
@staticmethod
def _recurse_apply(o: Any, fn: Callable[[Any], Any]) -> Any:
# TODO: dicts
# TODO: dict and set
if isinstance(o, (list, tuple)):
L = []
for item in o:
@ -2379,7 +2381,7 @@ class LazyTorchTensor:
def __add__(self, *args): # gemma
return self._wrap_fn(torch.Tensor.__add__)(self, *args)
def __getitem__(self, *args): # bloom falcon internlm2
def __getitem__(self, *args): # bloom falcon refact internlm2
return self._wrap_fn(torch.Tensor.__getitem__)(self, *args)