Skip to content

fix a bug in m3 modeling.py: #1410 #1424

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
hanhainebula merged 1 commit into from
Apr 10, 2025
Merged

fix a bug in m3 modeling.py: #1410 #1424

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
6 changes: 6 additions & 0 deletions FlagEmbedding/finetune/embedder/encoder_only/m3/modeling.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -422,6 +422,12 @@ def forward(
q_dense_vecs.size(0)*self.process_rank : q_dense_vecs.size(0)*(self.process_rank+1),
p_dense_vecs.size(0)*self.process_rank : p_dense_vecs.size(0)*(self.process_rank+1)
] # (batch_size, batch_size * group_size)
elif no_in_batch_neg_flag:
# get local p_dense_vecs: fix a bug described in
# https://github.com/FlagOpen/FlagEmbedding/issues/1410
group_size = p_dense_vecs.size(0) // q_dense_vecs.size(0)
indices = torch.arange(0, q_dense_vecs.size(0), device=q_dense_vecs.device) * group_size
p_dense_vecs = p_dense_vecs[indices, :]

# ensemble loss
ensemble_scores, ensemble_loss = compute_loss_func(
Expand Down
Loading