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

Optimize for topk=1 case if we do not handle duplicates #603

Open
wants to merge 2 commits into
base: habana_main
Choose a base branch
from
Open

Optimize for topk=1 case if we do not handle duplicates #603

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
208e21b
Optimize for topk=1 case if we do not handle duplicates
ssarkar2 Dec 9, 2024
c6023de
style
ssarkar2 Dec 11, 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
8 changes: 8 additions & 0 deletions vllm/model_executor/layers/sampler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -405,6 +405,14 @@ def __init__(self, increment: int):
self._increment = increment

def __call__(self, logits: torch.Tensor, p: float, k: int):
if k == 1 and not ApplyToppTopkScalar._handle_duplicates:
new_logits = torch.full(logits.shape,
-float("inf"),
device=logits.device)
vals, idx = torch.max(logits, keepdim=True, dim=1)
new_logits.scatter_(1, idx, vals.to(new_logits.dtype))
return new_logits

if k > ApplyToppTopkScalar._padded_k:
ApplyToppTopkScalar._padded_k = min(k + self._increment,
logits.shape[1])
Expand Down
Loading