Skip to content

[PyTorch] Share CUDA graph memory across dynamic CP variants #3353

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

Draft
xiaoyao0115 wants to merge 30 commits into
base: release_v2.14
Choose a base branch
from
+2,930 −102
Draft

[PyTorch] Share CUDA graph memory across dynamic CP variants #3353

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
30 commits
Select commit Hold shift + click to select a range
a9be8ee
Share CUDA graph memory across dynamic CP variants
xiaoyao0115 Aug 12, 2026
4a7b8fd
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Aug 12, 2026
c095214
Fix CUDA graph slot memory lint
xiaoyao0115 Aug 12, 2026
6047839
Fix dynamic-CP CUDA graph slot wrap
Aug 17, 2026
85e4dcd
Speed up CUDA graph slot owner lookup
Aug 19, 2026
9b48ea5
Harden CUDA graph slot capture lifecycle
Aug 21, 2026
8fa96bb
Fix allocator guard lint
Aug 21, 2026
209e757
Honor user gradient liveness in graph slots
Aug 21, 2026
6110d0a
Protect saved CUDA graph kwargs across slot schedules
Aug 21, 2026
ffef962
Roll back failed CUDA graph allocator restores
Aug 21, 2026
db1f42b
Roll back partial CUDA graph owner release
Aug 21, 2026
461d00f
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Aug 21, 2026
032b62f
Restore CUDA graph warmup hooks on failure
Aug 21, 2026
de59f96
Preserve aliased outputs in CUDA graph slots
Aug 21, 2026
b07d7df
Skip empty pipeline chunks in graph slot grouping
Aug 21, 2026
d12ff87
Harden CUDA graph slot prerequisites
Aug 22, 2026
28980c2
Fix typed CUDA graph arena offsets
Aug 22, 2026
9f4ec8a
Test CUDA graph staging input rebinding
Aug 22, 2026
7741fea
Preserve CUDA graph staging input aliases
Aug 22, 2026
6ca7a88
Reject incompatible CUDA graph output aliases
Aug 22, 2026
f799562
Reject CUDA graph inputs aliased to module state
Aug 22, 2026
0c71af0
Reject cross-callable CUDA graph state aliases
Aug 22, 2026
785c03c
Recheck CUDA graph state aliases after warmup
Aug 22, 2026
be515df
Reject non-Tensor CUDA graph slot kwargs
Aug 22, 2026
f5a36a9
Simplify CUDA graph slot checkpoint state
Aug 23, 2026
ff6697a
Remove CUDA graph allocator checkpointing
Aug 23, 2026
8a9f58c
Avoid duplicate CUDA graph saved-tensor spill
Aug 23, 2026
3c9a942
Simplify CUDA graph arena sizing
Aug 23, 2026
86ba060
Simplify CUDA graph slot arena placement
Aug 23, 2026
e2fc956
Move CP P2P transport override to a standalone change
Aug 24, 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
Loading
Loading