Skip to content
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

[Fix] Fix the case where prompt_len = 0 #1593

Merged
merged 2 commits into from
Oct 7, 2024
Merged

[Fix] Fix the case where prompt_len = 0 #1593

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
d574f0e
Fix the case where prompt_len = 0
merrymercy Oct 7, 2024
96f656f
update
merrymercy Oct 7, 2024
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: 4 additions & 2 deletions python/sglang/srt/managers/scheduler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -626,9 +626,11 @@ def run_batch(self, batch: ScheduleBatch):
else:
logits_output = None
if self.tokenizer is not None:
next_token_ids = [self.tokenizer.eos_token_id] * len(batch.reqs)
next_token_ids = torch.full(
(batch.batch_size(),), self.tokenizer.eos_token_id
)
else:
next_token_ids = [0] * len(batch.reqs)
next_token_ids = torch.full((batch.batch_size(),), 0)
return logits_output, next_token_ids
else: # embedding or reward model
assert batch.extend_num_tokens != 0
Expand Down
2 changes: 1 addition & 1 deletion python/sglang/srt/managers/tokenizer_manager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -526,7 +526,7 @@ async def update_weights(
async with self.model_update_lock:
# wait for the previous generation requests to finish
while len(self.rid_to_state) > 0:
await asyncio.sleep(0)
await asyncio.sleep(0.001)
self.send_to_scheduler.send_pyobj(obj)
self.model_update_result = asyncio.Future()
result = await self.model_update_result
Expand Down
2 changes: 1 addition & 1 deletion python/sglang/srt/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -624,6 +624,6 @@ def broadcast_pyobj(
tensor_data = torch.empty(size, dtype=torch.uint8)
dist.broadcast(tensor_data, src=0, group=dist_group)

serialized_data = bytes(tensor_data.tolist())
serialized_data = bytes(tensor_data.cpu().numpy())
data = pickle.loads(serialized_data)
return data
Loading