Skip to content

convert : workaround for AutoConfig dummy labels #13881

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

Merged
merged 1 commit into from
May 29, 2025
Merged

convert : workaround for AutoConfig dummy labels #13881

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
12 changes: 9 additions & 3 deletions convert_hf_to_gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3690,14 +3690,20 @@ def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.vocab_size = None

if cls_out_labels := self.hparams.get("id2label"):
if len(cls_out_labels) == 2 and cls_out_labels[0] == "LABEL_0":
# Remove dummy labels added by AutoConfig
cls_out_labels = None
self.cls_out_labels = cls_out_labels

def set_gguf_parameters(self):
super().set_gguf_parameters()
self.gguf_writer.add_causal_attention(False)
self._try_set_pooling_type()

if cls_out_labels := self.hparams.get("id2label"):
if self.cls_out_labels:
key_name = gguf.Keys.Classifier.OUTPUT_LABELS.format(arch = gguf.MODEL_ARCH_NAMES[self.model_arch])
self.gguf_writer.add_array(key_name, [v for k, v in sorted(cls_out_labels.items())])
self.gguf_writer.add_array(key_name, [v for k, v in sorted(self.cls_out_labels.items())])

def set_vocab(self):
tokens, toktypes, tokpre = self.get_vocab_base()
Expand Down Expand Up @@ -3749,7 +3755,7 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
if name.startswith("cls.seq_relationship"):
return []

if self.hparams.get("id2label"):
if self.cls_out_labels:
# For BertForSequenceClassification (direct projection layer)
if name == "classifier.weight":
name = "classifier.out_proj.weight"
Expand Down