Skip to content

nixl pd support qwen3.5 #1340

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
hiworldwzj merged 3 commits into from
Jun 9, 2026
Merged

nixl pd support qwen3.5 #1340

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c0bdab0
nixl pd support qwen3.5
hiworldwzj Jun 9, 2026
d6bf28b
fix
hiworldwzj Jun 9, 2026
7999513
fix
hiworldwzj Jun 9, 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: 6 additions & 0 deletions lightllm/common/kv_cache_mem_manager/deepseek2_mem_manager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,10 @@ def write_mem_to_page_kv_move_buffer(
dp_index: int,
mem_managers: List["MemoryManager"],
dp_world_size: int,
page_kind: str = "kv",
req_idx: int = None,
):
assert page_kind == "kv", f"{type(self).__name__} does not support page_kind={page_kind}"
cur_page = self.kv_move_buffer[page_index]
pin_mem_indexes = self._buffer_mem_indexes_tensors[page_index][0 : len(mem_indexes)]
pin_mem_indexes.numpy()[:] = mem_indexes
Expand All @@ -77,7 +80,10 @@ def read_page_kv_move_buffer_to_mem(
dp_index: int,
mem_managers: List["MemoryManager"],
dp_world_size: int,
page_kind: str = "kv",
req_idx: int = None,
):
assert page_kind == "kv", f"{type(self).__name__} does not support page_kind={page_kind}"
cur_page = self.kv_move_buffer[page_index]
pin_mem_indexes = self._buffer_mem_indexes_tensors[page_index][0 : len(mem_indexes)]
pin_mem_indexes.numpy()[:] = mem_indexes
Expand Down
9 changes: 6 additions & 3 deletions lightllm/common/kv_cache_mem_manager/mem_manager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,9 +102,6 @@ def alloc_kv_move_buffer(self, max_req_total_len):
return

def alloc_paged_kv_move_buffer(self, page_num, page_size) -> torch.Tensor:
if isinstance(self, MemoryManager) and type(self) is not MemoryManager:
raise NotImplementedError("subclass need reimpl this method")

num_kv_head = get_num_key_value_heads(get_env_start_args().model_dir)
self.kv_move_buffer = torch.empty(
(page_num, page_size, self.layer_num, 2 * num_kv_head, self.head_dim), dtype=self.dtype, device="cuda"
Expand All @@ -121,7 +118,10 @@ def write_mem_to_page_kv_move_buffer(
dp_index: int,
mem_managers: List["MemoryManager"],
dp_world_size: int,
page_kind: str = "kv",
req_idx: int = None,
):
assert page_kind == "kv", f"{type(self).__name__} does not support page_kind={page_kind}"
cur_page = self.kv_move_buffer[page_index]
pin_mem_indexes = self._buffer_mem_indexes_tensors[page_index][0 : len(mem_indexes)]
pin_mem_indexes.numpy()[:] = mem_indexes
Expand Down Expand Up @@ -150,7 +150,10 @@ def read_page_kv_move_buffer_to_mem(
dp_index: int,
mem_managers: List["MemoryManager"],
dp_world_size: int,
page_kind: str = "kv",
req_idx: int = None,
):
assert page_kind == "kv", f"{type(self).__name__} does not support page_kind={page_kind}"
cur_page = self.kv_move_buffer[page_index]
pin_mem_indexes = self._buffer_mem_indexes_tensors[page_index][0 : len(mem_indexes)]
pin_mem_indexes.numpy()[:] = mem_indexes
Expand Down
Loading
Loading