Set krea 2 to use new lokr format
This commit is contained in:
@@ -157,6 +157,7 @@ class Krea2Model(BaseModel):
|
||||
# Qwen2TokenizerFast used to tokenize the assistant suffix (matches the
|
||||
# reference's separate processor pass).
|
||||
self.processor = None
|
||||
self.use_old_lokr_format = False
|
||||
|
||||
@staticmethod
|
||||
def get_train_scheduler():
|
||||
|
||||
Reference in New Issue
Block a user