Skip to content

[Fix][Qwen3.5] Pass max_mamba_cache_size to mamba pool in disaggregation decode path #19002

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
YAMY1234 wants to merge 2 commits into
base: main
Choose a base branch
from
+6 −1
Open

[Fix][Qwen3.5] Pass max_mamba_cache_size to mamba pool in disaggregation decode path #19002

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
0270147
set mamba_size for HybridMambaDecodeReqToTokenPool
Feb 19, 2026
19a07b5
format
Feb 19, 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
6 changes: 5 additions & 1 deletion python/sglang/srt/disaggregation/decode.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -158,6 +158,7 @@ def __init__(
speculative_num_draft_tokens: int,
enable_mamba_extra_buffer: bool,
pre_alloc_size: int,
mamba_size: int = None,
Copy link
Collaborator

@ShangmingCai ShangmingCai Feb 19, 2026

Choose a reason for hiding this comment

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

It is weird that we have a size and a mamba size for HybridMambaDecodeReqToTokenPool at the same time.

Copy link
Contributor Author

@YAMY1234 YAMY1234 Feb 19, 2026

Choose a reason for hiding this comment

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

I had similar concern when I was trying to understand this, but I found its parent class HybridReqToTokenPool also takes both size and mamba_size as separate parameters — they represent different things (max_num_reqs for req-to-token vs max_mamba_cache_size for mamba state)

):
DecodeReqToTokenPool.__init__(
self,
Expand All @@ -172,8 +173,11 @@ def __init__(
)
self.enable_mamba_extra_buffer = enable_mamba_extra_buffer
self.enable_memory_saver = enable_memory_saver
effective_mamba_size = (
mamba_size if mamba_size is not None else size
) + pre_alloc_size
self._init_mamba_pool(
size=size + pre_alloc_size,
size=effective_mamba_size,
mamba_spec_state_size=size + pre_alloc_size,
cache_params=cache_params,
device=device,
Expand Down
1 change: 1 addition & 0 deletions python/sglang/srt/model_executor/model_runner_kv_cache_mixin.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -445,6 +445,7 @@ def init_memory_pool(self: ModelRunner, total_gpu_memory: int):
speculative_num_draft_tokens=self.server_args.speculative_num_draft_tokens,
enable_mamba_extra_buffer=self.server_args.enable_mamba_extra_buffer(),
pre_alloc_size=pre_alloc_size,
mamba_size=self.server_args.max_mamba_cache_size,
)
else:
self.req_to_token_pool = DecodeReqToTokenPool(
Expand Down
Loading