Skip to content

Fix/dsv4 cp relax assert offline packed thd #6158

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
thomasdhc merged 2 commits into from
Jul 31, 2026
+0 −7
Merged

Fix/dsv4 cp relax assert offline packed thd #6158

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
10b6f90
fix(dsv4-cp): relax sequence_packing_scheduler assert for offline-pac…
weijiac0619 Jul 27, 2026
efc4009
fix(dsv4-cp): remove sequence_packing_scheduler assert for DSv4 hybri…
weijiac0619 Jul 30, 2026
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: 0 additions & 7 deletions megatron/core/transformer/transformer_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1531,13 +1531,6 @@ def __post_init__(self):
if self.cp_partition_mode not in ("zigzag", "contiguous"):
raise ValueError(f"Unsupported cp_partition_mode: {self.cp_partition_mode}")

if self.experimental_attention_variant == "dsv4_hybrid" and (
self.context_parallel_size > 1 or self.dynamic_context_parallel
):
assert (
self.sequence_packing_scheduler is not None
), "DSv4 Hybrid with CP requires a sequence_packing_scheduler for THD inputs."

if self.context_parallel_size > 1:
if (
self.experimental_attention_variant == "dsv4_hybrid"
Expand Down
Loading