Skip to content

Please enter the commit message for your changes. Lines starting #273

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kmransom56 wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Please enter the commit message for your changes. Lines starting #273

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
faf6af9
Please enter the commit message for your changes. Lines starting
kmransom56 Mar 7, 2026
5d72c6b
Update gguf-py/gguf/gguf_reader.py
kmransom56 Mar 7, 2026
69786f1
Harden webui localStorage parsing and migration validation
kmransom56 Mar 10, 2026
4ce203d
modified: smallthinker/docs/docker.md
kmransom56 Mar 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
808 changes: 553 additions & 255 deletions convert-dense.py
View file
Open in desktop

Large diffs are not rendered by default.

28 changes: 17 additions & 11 deletions convert-hf-to-powerinfer-gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,7 +90,10 @@ def __init__(
self.hparams = Model.load_hparams(self.dir_model)
self.model_arch = self._get_model_architecture()
self.gguf_writer = gguf.GGUFWriter(
fname_out, gguf.MODEL_ARCH_NAMES[self.model_arch], endianess=self.endianess, use_temp_file = False
fname_out,
gguf.MODEL_ARCH_NAMES[self.model_arch],
endianess=self.endianess,
use_temp_file=False,
)

def set_vocab(self):
Expand Down Expand Up @@ -517,6 +520,7 @@ def write_tensors(self):

self.gguf_writer.add_tensor(new_name, data)


class OptModel(Model):
def set_gguf_parameters(self, params: PredictorParams):
self.gguf_writer.add_name("opt")
Expand All @@ -527,20 +531,20 @@ def set_gguf_parameters(self, params: PredictorParams):
self.gguf_writer.add_head_count(self.hparams["num_attention_heads"])
# self.gguf_writer.add_vocab_size(self.hparams["vocab_size"])
self.gguf_writer.add_file_type(self.ftype)

if params.sparse_threshold is not None:
self.gguf_writer.add_sparse_threshold(params.sparse_threshold)

def write_tensors(self):
for name, data_torch in self.get_tensors():
old_dtype = data_torch.dtype

# convert any unsupported data types to float32
if data_torch.dtype not in (torch.float16, torch.float32):
data_torch = data_torch.to(torch.float32)

data = data_torch.squeeze().numpy()

# map tensor names
new_name = self._translate_tensor_key(name)
if new_name is None:
Expand All @@ -552,8 +556,8 @@ def write_tensors(self):
if "ffn_down" in new_name:
new_name = new_name.replace("ffn_down", "ffn_down_t")
data = data.T
n_dims = len(data.shape)

n_dims = len(data.shape)
data_dtype = data.dtype

# if f32 desired, convert any float16 to float32
Expand All @@ -570,11 +574,12 @@ def write_tensors(self):
and n_dims == 2
):
data = data.astype(np.float16)

print(f"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}")

self.gguf_writer.add_tensor(new_name, data)


@dataclass
class PredictorParams:
sparse_threshold: float | None = None
Expand All @@ -583,12 +588,12 @@ class PredictorParams:
def loadPredictorJson(config_path: Path) -> PredictorParams:
config = json.load(open(config_path))
return PredictorParams(
sparse_threshold = config.get("sparse_threshold"),
sparse_threshold=config.get("sparse_threshold"),
)

@staticmethod
def load(model_instance: Model) -> PredictorParams:
config_path = model_instance.dir_mlp_pred / "config.json"
config_path = model_instance.dir_mlp_pred / "config.json"

if config_path.exists():
params = PredictorParams.loadPredictorJson(config_path)
Expand All @@ -597,6 +602,7 @@ def load(model_instance: Model) -> PredictorParams:

return params


###### CONVERSION LOGIC ######


Expand Down
Loading