Skip to content

[BugFix] Decoupling enable_mm and thinking binding #4314

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
wants to merge 2 commits into
base: feature/experimental_feature_20250908
Choose a base branch
from
Open

[BugFix] Decoupling enable_mm and thinking binding #4314

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
69e6685
[BugFix] Decoupling enable_mm and thinking binding
CSWYF3634076 Sep 29, 2025
39d60ae
[Test] add qwen25 vl unit testing
CSWYF3634076 Sep 29, 2025
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
44 changes: 34 additions & 10 deletions fastdeploy/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -264,15 +264,27 @@ def insert_tasks_v1(self, req_dicts: List[Request], num_running_requests: int =
else:
position_ids = None

enable_thinking = request.get("enable_thinking", True)
enable_thinking = enable_thinking if enable_thinking is not None else True
self.share_inputs["enable_thinking"][:] = enable_thinking
self.share_inputs["need_think_end"][idx : idx + 1, :] = 1 if enable_thinking else 0
self.share_inputs["reasoning_index"][idx : idx + 1, :] = request.get("reasoning_max_tokens", 2048)
self.share_inputs["rope_emb"][idx : idx + 1, :] = self.prepare_rope3d(
position_ids, request.get("max_tokens", 2048)
)

if request.get("enable_thinking", False):
# Enable thinking
req_reasoning_max_tokens = request.get("reasoning_max_tokens")
req_max_tokens = request.get("max_tokens")
final_reasoning_tokens = (
req_reasoning_max_tokens if req_reasoning_max_tokens is not None else req_max_tokens
)

self.share_inputs["enable_thinking"][idx : idx + 1] = True
self.share_inputs["need_think_end"][idx : idx + 1, :] = 1
self.share_inputs["reasoning_index"][idx : idx + 1, :] = final_reasoning_tokens
else:
# Disable thinking
self.share_inputs["enable_thinking"][idx : idx + 1] = False
self.share_inputs["need_think_end"][idx : idx + 1, :] = 0
self.share_inputs["reasoning_index"][idx : idx + 1, :] = 0

if isinstance(request.prompt_token_ids, np.ndarray):
prompt_token_ids = request.prompt_token_ids.tolist()
else:
Expand Down Expand Up @@ -496,16 +508,28 @@ def insert_prefill_inputs(self, req_dicts: List[Request], num_running_requests:
self.share_inputs["prompt_lens"][idx : idx + 1] = length

if self.enable_mm:
enable_thinking = request.get("enable_thinking", True)
enable_thinking = enable_thinking if enable_thinking is not None else True
self.share_inputs["enable_thinking"][:] = enable_thinking
self.share_inputs["need_think_end"][idx : idx + 1, :] = 1 if enable_thinking else 0
self.share_inputs["reasoning_index"][idx : idx + 1, :] = request.get("reasoning_max_tokens", 2048)
self.share_inputs["rope_emb"][idx : idx + 1, :] = self.prepare_rope3d(
position_ids, request.get("max_tokens", 2048)
)
self.share_inputs["seq_lens_decoder"][idx : idx + 1] = 0

if request.get("enable_thinking", False):
# Enable thinking
req_reasoning_max_tokens = request.get("reasoning_max_tokens")
req_max_tokens = request.get("max_tokens")
final_reasoning_tokens = (
req_reasoning_max_tokens if req_reasoning_max_tokens is not None else req_max_tokens
)

self.share_inputs["enable_thinking"][idx : idx + 1] = True
self.share_inputs["need_think_end"][idx : idx + 1, :] = 1
self.share_inputs["reasoning_index"][idx : idx + 1, :] = final_reasoning_tokens
else:
# Disable thinking
self.share_inputs["enable_thinking"][idx : idx + 1] = False
self.share_inputs["need_think_end"][idx : idx + 1, :] = 0
self.share_inputs["reasoning_index"][idx : idx + 1, :] = 0

def get_attr_from_request(request, attr, default_value=None):
res = request.get(attr, default_value)
if res is not None:
Expand Down
Loading
Loading