Skip to content

fix: remove invalid deepstack boundary check for Qwen3-VL #1664

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
AlpinDale merged 1 commit into from
May 1, 2026
Merged

fix: remove invalid deepstack boundary check for Qwen3-VL #1664

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
11 changes: 0 additions & 11 deletions aphrodite/model_executor/models/qwen3_omni_moe_thinker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1666,11 +1666,6 @@ def _get_deepstack_input_embeds(
return None # If vision tower is skipped
if getattr(self, "deepstack_input_embeds_num_tokens", 0) == 0:
return None
if num_tokens > self.deepstack_input_embeds_num_tokens:
raise ValueError(
"Requested more deepstack tokens than available in buffer: "
f"{num_tokens=} > {self.deepstack_input_embeds_num_tokens=}"
)

# get deepstack_input_embeds from buffer, and clear the buffer
return IntermediateTensors(
Expand Down Expand Up @@ -1709,12 +1704,6 @@ def _clear_deepstack_input_embeds(self, num_tokens: int) -> None:

# clear deepstack_input_embeds in buffer
if num_tokens > 0:
if num_tokens > self.deepstack_input_embeds_num_tokens:
raise ValueError(
"Requested to clear more deepstack tokens than available in "
"buffer: "
f"{num_tokens=} > {self.deepstack_input_embeds_num_tokens=}"
)
for idx in range(self.deepstack_num_level):
self.deepstack_input_embeds[idx][:num_tokens].zero_()
self.deepstack_input_embeds_num_tokens = 0
Expand Down
11 changes: 0 additions & 11 deletions aphrodite/model_executor/models/qwen3_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1624,11 +1624,6 @@ def _get_deepstack_input_embeds(
return None # If vision tower is skipped
if getattr(self, "deepstack_input_embeds_num_tokens", 0) == 0:
return None
if num_tokens > self.deepstack_input_embeds_num_tokens:
raise ValueError(
"Requested more deepstack tokens than available in buffer: "
f"{num_tokens=} > {self.deepstack_input_embeds_num_tokens=}"
)

# get deepstack_input_embeds from buffer, and clear the buffer
return IntermediateTensors(
Expand Down Expand Up @@ -1666,12 +1661,6 @@ def _clear_deepstack_input_embeds(self, num_tokens: int) -> None:

# clear deepstack_input_embeds in buffer
if num_tokens > 0:
if num_tokens > self.deepstack_input_embeds_num_tokens:
raise ValueError(
"Requested to clear more deepstack tokens than available in "
"buffer: "
f"{num_tokens=} > {self.deepstack_input_embeds_num_tokens=}"
)
for idx in range(self.deepstack_num_level):
self.deepstack_input_embeds[idx][:num_tokens].zero_()
self.deepstack_input_embeds_num_tokens = 0
Expand Down
Loading