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

Proper check for None type for new_name to avoid crash; formatting; revert change to base class write_tensors()

Browse source
This commit is contained in:
Ashish 2024-04-14 14:28:12 -07:00
parent 96695fb96b
commit 13c75c21eb
1 changed files with 5 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

9
convert-hf-to-gguf.py
View file

@ -164,7 +164,7 @@ class Model(ABC):
data = data.astype(np.float32) data = data.astype(np.float32)
# if f16 desired, convert any float32 2-dim weight tensors to float16 # if f16 desired, convert any float32 2-dim weight tensors to float16
if self.ftype == 1 and data_dtype == np.float32 and name.endswith(".weight") and not new_name.endswith("_norm.weight") and n_dims == 2: if self.ftype == 1 and data_dtype == np.float32 and name.endswith(".weight") and n_dims == 2:
data = data.astype(np.float16) data = data.astype(np.float16)
print(f"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}") print(f"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}")
@ -1264,6 +1264,7 @@ class StableLMModel(Model):
print(f"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}") print(f"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}")
self.gguf_writer.add_tensor(new_name, data) self.gguf_writer.add_tensor(new_name, data)
def _stack_qk_norm(self, block_count, name, tensor_map, n_head, norms, n_dims, layer_name="q_layernorm"): def _stack_qk_norm(self, block_count, name, tensor_map, n_head, norms, n_dims, layer_name="q_layernorm"):
for bid in range(block_count): for bid in range(block_count):
datas = [] datas = []
@ -1275,15 +1276,15 @@ class StableLMModel(Model):
data_dtype = data.dtype data_dtype = data.dtype
merged_name = f"model.layers.{bid}.self_attn.{layer_name}.weight" merged_name = f"model.layers.{bid}.self_attn.{layer_name}.weight"
new_name = tensor_map.get_name(merged_name, try_suffixes=(".weight", ".bias")) new_name = tensor_map.get_name(merged_name, try_suffixes=(".weight", ".bias"))
if new_name is None:
print(f"Can not map tensor {name!r}")
sys.exit()
if self.ftype == 1 and data_dtype == np.float16 and (n_dims == 1 or new_name.endswith("_norm.weight")): if self.ftype == 1 and data_dtype == np.float16 and (n_dims == 1 or new_name.endswith("_norm.weight")):
data = data.astype(np.float32) data = data.astype(np.float32)
# if f16 desired, convert any float32 2-dim weight tensors to float16 # if f16 desired, convert any float32 2-dim weight tensors to float16
if self.ftype == 1 and data_dtype == np.float32 and name.endswith(".weight") and not new_name.endswith("_norm.weight") and n_dims == 2: if self.ftype == 1 and data_dtype == np.float32 and name.endswith(".weight") and not new_name.endswith("_norm.weight") and n_dims == 2:
data = data.astype(np.float16) data = data.astype(np.float16)
if new_name is None:
print(f"Can not map tensor {name!r}")
sys.exit()
print(f"{new_name}, n_dims = {len(data.shape)}, shape = {data.shape} --> {data.dtype}") print(f"{new_name}, n_dims = {len(data.shape)}, shape = {data.shape} --> {data.dtype}")