gguf : enforce that tensor names are unique (#6905)

* not allow adding duplicated tensor name

* no duplicated tensor while reading gguf

* typo

* throw exception inside llama_model_loader

Co-authored-by: slaren <slarengh@gmail.com>

---------

Co-authored-by: slaren <slarengh@gmail.com>
This commit is contained in:
Xuan Son Nguyen
2024-04-28 17:36:18 +02:00
committed by GitHub
parent ce023f6f2f
commit 7bb36ccf91
4 changed files with 32 additions and 1 deletions

View File

@ -63,6 +63,7 @@ class GGUFWriter:
self.kv_data_count = 0
self.ti_data = bytearray()
self.ti_data_count = 0
self.ti_names = set()
self.use_temp_file = use_temp_file
self.temp_file = None
self.tensors = []
@ -197,6 +198,10 @@ class GGUFWriter:
if self.state is not WriterState.EMPTY:
raise ValueError(f'Expected output file to be empty, got {self.state}')
if name in self.ti_names:
raise ValueError(f'Duplicated tensor name {name}')
self.ti_names.add(name)
encoded_name = name.encode("utf8")
self.ti_data += self._pack("Q", len(encoded_name))
self.ti_data += encoded_name