Skip to content

Bug in DenoisingAutoEncoderLoss.py #2619

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
tomaarsen merged 3 commits into from
May 8, 2024
Merged

Bug in DenoisingAutoEncoderLoss.py #2619

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
14b07fb
Update DenoisingAutoEncoderLoss.py
arun477 Apr 27, 2024
64c0dc8
compatibility for transformers >=4.40.0 added
arun477 May 1, 2024
52746d6
format issue updated
arun477 May 1, 2024
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
18 changes: 15 additions & 3 deletions sentence_transformers/losses/DenoisingAutoEncoderLoss.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,9 +113,21 @@ def __init__(self, model: SentenceTransformer, decoder_name_or_path: str = None,
"Since the encoder vocabulary has been changed and --tie_encoder_decoder=True, now the new vocabulary has also been used for the decoder."
)
decoder_base_model_prefix = self.decoder.base_model_prefix
PreTrainedModel._tie_encoder_decoder_weights(
model[0].auto_model, self.decoder._modules[decoder_base_model_prefix], self.decoder.base_model_prefix
)
try:
# Compatibility with transformers <4.40.0
PreTrainedModel._tie_encoder_decoder_weights(
model[0].auto_model,
self.decoder._modules[decoder_base_model_prefix],
self.decoder.base_model_prefix,
)
except TypeError:
# Compatibility with transformers >=4.40.0
PreTrainedModel._tie_encoder_decoder_weights(
model[0].auto_model,
self.decoder._modules[decoder_base_model_prefix],
self.decoder.base_model_prefix,
encoder_name_or_path,
)

def retokenize(self, sentence_features):
input_ids = sentence_features["input_ids"]
Expand Down