Skip to content

Ring Attention: free unused kv comm buffers #3411

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
francesco-bertolotti wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −0
Open

Ring Attention: free unused kv comm buffers #3411

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 transformer_engine/pytorch/attention/dot_product_attention/context_parallel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1715,6 +1715,12 @@ def forward(
)

kv_inputs[i % 2] = p2p_comm_buffers[i]

# Release the KV chunk from two steps back.
if i >= 2 and not is_graph_capturing():
p2p_comm_buffers[i - 2].record_stream(flash_attn_streams[i % 2])
Comment on lines +1719 to +1721

@greptile-apps greptile-apps Bot Aug 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Cover the active eviction branch

The new buffer release executes only for pure P2P Ring Attention with a context-parallel size of at least three, but existing coverage does not exercise that configuration. Add a regression test for this branch so stream-lifetime and ring-index regressions do not leave larger runs vulnerable to corrupted results or GPU memory errors.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

p2p_comm_buffers[i - 2] = None

k_part = kv_inputs[i % 2][:k_numel].view(*k_shape)
v_part = kv_inputs[i % 2][k_numel:].view(*v_shape)
q_part = q
Expand Down
Loading