Skip to content

[Bugfix][Kimi-K3] Refresh DSpark context KV cache pointers after the KV cache is re-bound #58814

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
AndreasKaratzas merged 4 commits into from
Sep 28, 2026
+6 −1
Merged

[Bugfix][Kimi-K3] Refresh DSpark context KV cache pointers after the KV cache is re-bound #58814

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c4b5280
[Bugfix][Kimi-K3] Refresh DSpark context KV cache pointers after the …
okorzh-amd Sep 26, 2026
92b6ac9
Merge branch 'main' into okorzh/fix-k3-dspark-stale-kv-cache-ptrs
okorzh-amd Sep 26, 2026
d44cf80
Invalidate the cached context KV pointers instead of updating them in…
okorzh-amd Sep 26, 2026
73ae7a8
Merge branch 'main' into okorzh/fix-k3-dspark-stale-kv-cache-ptrs
okorzh-amd Sep 26, 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: 6 additions & 1 deletion vllm/models/kimi_k3/nvidia/dspark_mla.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -376,7 +376,12 @@ def _has_uniform_block_layout(
self,
cache_layers: list[MultiHeadLatentAttention],
) -> bool:
if not hasattr(self, "_layers_share_kv_block_layout"):
key = tuple(cl.kv_cache.data_ptr() for cl in cache_layers)

@Juntian777 Juntian777 Sep 26, 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.

Could we just invalidate both caches here when the kv_cache.data_ptr() tuple changes? Since the profiling graphs are discarded and the runtime graphs are recaptured after the real KV cache is bound, it seems like we don't need the in-place copy_() here. Then _get_context_kv_cache_ptrs() can stay unchanged.

@okorzh-amd okorzh-amd Sep 26, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

good point. I've updated the code.

if getattr(self, "_kv_cache_ptrs_key", None) != key:
assert not torch.cuda.is_current_stream_capturing()
self._kv_cache_ptrs_key = key
if hasattr(self, "_context_cache_ptrs"):
del self._context_cache_ptrs
ref_cache = cache_layers[0].kv_cache
self._layers_share_kv_block_layout = all(
cl.kv_cache.size(1) == ref_cache.size(1)
Expand Down
Loading