Skip to content

Fix off-by-one crash in forward_sliding1 for T ≡ 1 (mod chunk_size) #6

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
jiah-cloud merged 1 commit into from
Jul 3, 2026
+10 −0
Merged

Fix off-by-one crash in forward_sliding1 for T ≡ 1 (mod chunk_size) #6

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
10 changes: 10 additions & 0 deletions track4world/nets/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1409,6 +1409,16 @@ def forward_sliding1(
del fmaps_chunk, ctxfeats_chunk, pms_chunk, fmaps3d_detail_chunk
torch.cuda.empty_cache()

# The loop runs num_chunks = ceil((T-1)/chunk_size) times, but points/masks/
# world_points/camera_poses are split into ceil(T/chunk_size) chunks. When
# T ≡ 1 (mod chunk_size) the last chunk is dropped, leaving points one frame
# short (causes a shape mismatch in infer_pair). Append the leftover chunks.
for j in range(num_chunks, len(points_chunks)):
points_list.append(padder.unpad(points_chunks[j]))
masks_list.append(padder.unpad(masks_chunks[j]))
world_points_list.append(padder.unpad(world_points_chunks[j]))
camera_poses_list.append(camera_poses_chunks[j])

# --- Aggregate Chunks into Final Tensors ---
all_flow_preds, all_visconf_preds = [], []
all_flow3d_preds = [] if tracking3d else None
Expand Down