Skip to content

[vllm] fix: reset all caches after weight updates #6522

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
s-isaev wants to merge 2 commits into
base: main
Choose a base branch
from
+6 −1
Open

[vllm] fix: reset all caches after weight updates #6522

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4b5a1ae
[vllm] reset all caches after weight updates
s-isaev May 28, 2026
8fcb2f6
[vllm] address cache reset review comment
s-isaev May 29, 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
5 changes: 5 additions & 0 deletions verl/workers/rollout/vllm_rollout/vllm_async_server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -641,6 +641,11 @@ async def clear_kv_cache(self):
# is a no-op success, so we can pass it unconditionally.
await self.engine.reset_prefix_cache(**_RESET_PREFIX_CACHE_KWARGS)

if _VLLM_VERSION >= version.parse("0.9.0"):
await self.engine.reset_mm_cache()
Comment thread
s-isaev marked this conversation as resolved.
if _VLLM_VERSION >= version.parse("0.16.0"):
await self.engine.reset_encoder_cache()

async def release_kv_cache(self):
"""Release only kv_cache GPU memory, keeping model weights intact.
# TODO: support true release of kv_cache
Expand Down
2 changes: 1 addition & 1 deletion verl/workers/rollout/vllm_rollout/vllm_rollout.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,7 +190,7 @@ async def update_weights(
if future is not None:
await future

# reset prefix cache after updating weights
# reset caches after updating weights
if self.rollout_rank == 0:
await self.server_handle.clear_kv_cache.remote()
if global_steps is not None:
Expand Down