Skip to content

Unpack DiT block output tuple in the non-checkpointed training branch #86

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

Open
jeqcho wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Unpack DiT block output tuple in the non-checkpointed training branch #86

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
7 changes: 6 additions & 1 deletion groot/vla/model/dreamzero/modules/wan_video_dit_action_casual_chunk.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2140,7 +2140,12 @@ def custom_forward(*inputs, **kwargs):
use_reentrant=False,
)
else:
x = block(x, **kwargs)
# Blocks return (x, kv_cache); only the checkpointed branch
# unpacked it, so any grad-disabled training forward (e.g.
# deterministic eval of the training objective) crashed with
# AttributeError: 'tuple' object has no attribute 'shape'.
x, _updated_kv_cache = block(x, **kwargs)
assert _updated_kv_cache is None

if clean_x is not None:
x = x[:, clean_x.shape[1]:]
Expand Down