From e3bd93655db20ed246417a9ad7606777360f5129 Mon Sep 17 00:00:00 2001 From: Aleksei Nikiforov Date: Thu, 27 Nov 2025 14:12:51 +0100 Subject: [PATCH] Fix gguf_new_metadata.py and gguf_editor_gui.py for non-native endian files gguf_new_metadata.py reads data from reader. Reader doesn't byteswap tensors to native endianness. But writer does expect tensors in native endianness to convert them into requested endianness. There are two ways to fix this: update reader and do conversion to native endianness and back, or skip converting endianness in writer in this particular USE-case. gguf_editor_gui.py doesn't allow editing or viewing tensor data. Let's go with skipping excessive byteswapping. If eventually capability to view or edit tensor data is added, tensor data should be instead byteswapped when reading it. --- gguf-py/gguf/gguf_writer.py | 18 ++++++++++++------ gguf-py/gguf/scripts/gguf_editor_gui.py | 2 +- gguf-py/gguf/scripts/gguf_new_metadata.py | 2 +- 3 files changed, 14 insertions(+), 8 deletions(-) diff --git a/gguf-py/gguf/gguf_writer.py b/gguf-py/gguf/gguf_writer.py index 57ca2035fe2..8ddd895cb7c 100644 --- a/gguf-py/gguf/gguf_writer.py +++ b/gguf-py/gguf/gguf_writer.py @@ -371,10 +371,13 @@ def add_tensor_info( def add_tensor( self, name: str, tensor: np.ndarray[Any, Any], raw_shape: Sequence[int] | None = None, - raw_dtype: GGMLQuantizationType | None = None, + raw_dtype: GGMLQuantizationType | None = None, tensor_endianess: GGUFEndian | None = None ) -> None: - if (self.endianess == GGUFEndian.BIG and sys.byteorder != 'big') or \ - (self.endianess == GGUFEndian.LITTLE and sys.byteorder != 'little'): + # if tensor endianness is not passed, assume it's native to system + if tensor_endianess is None: + tensor_endianess = GGUFEndian.BIG if sys.byteorder == 'big' else GGUFEndian.LITTLE + + if tensor_endianess != self.endianess: # Don't byteswap inplace since lazy copies cannot handle it tensor = tensor.byteswap(inplace=False) if self.use_temp_file and self.temp_file is None: @@ -397,13 +400,16 @@ def write_padding(self, fp: IO[bytes], n: int, align: int | None = None) -> None if pad != 0: fp.write(bytes([0] * pad)) - def write_tensor_data(self, tensor: np.ndarray[Any, Any]) -> None: + def write_tensor_data(self, tensor: np.ndarray[Any, Any], tensor_endianess: GGUFEndian | None = None) -> None: if self.state is not WriterState.TI_DATA and self.state is not WriterState.WEIGHTS: raise ValueError(f'Expected output file to contain tensor info or weights, got {self.state}') assert self.fout is not None - if (self.endianess == GGUFEndian.BIG and sys.byteorder != 'big') or \ - (self.endianess == GGUFEndian.LITTLE and sys.byteorder != 'little'): + # if tensor endianness is not passed, assume it's native to system + if tensor_endianess is None: + tensor_endianess = GGUFEndian.BIG if sys.byteorder == 'big' else GGUFEndian.LITTLE + + if tensor_endianess != self.endianess: # Don't byteswap inplace since lazy copies cannot handle it tensor = tensor.byteswap(inplace=False) diff --git a/gguf-py/gguf/scripts/gguf_editor_gui.py b/gguf-py/gguf/scripts/gguf_editor_gui.py index 05f4db0f8cd..293316afed4 100755 --- a/gguf-py/gguf/scripts/gguf_editor_gui.py +++ b/gguf-py/gguf/scripts/gguf_editor_gui.py @@ -1552,7 +1552,7 @@ def save_file(self): # Add tensors (including data) for tensor in self.reader.tensors: - writer.add_tensor(tensor.name, tensor.data, raw_shape=tensor.data.shape, raw_dtype=tensor.tensor_type) + writer.add_tensor(tensor.name, tensor.data, raw_shape=tensor.data.shape, raw_dtype=tensor.tensor_type, tensor_endianess=self.reader.endianess) # Write header and metadata writer.open_output_file(Path(file_path)) diff --git a/gguf-py/gguf/scripts/gguf_new_metadata.py b/gguf-py/gguf/scripts/gguf_new_metadata.py index 2fa5800cf74..c67436bad4a 100755 --- a/gguf-py/gguf/scripts/gguf_new_metadata.py +++ b/gguf-py/gguf/scripts/gguf_new_metadata.py @@ -94,7 +94,7 @@ def copy_with_new_metadata(reader: gguf.GGUFReader, writer: gguf.GGUFWriter, new writer.write_ti_data_to_file() for tensor in reader.tensors: - writer.write_tensor_data(tensor.data) + writer.write_tensor_data(tensor.data, tensor_endianess=reader.endianess) bar.update(tensor.n_bytes) writer.close()