Skip to content

Add mask_val option to StaticAttentionmask #8736

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
merged 1 commit into from
Feb 26, 2025
Merged

Add mask_val option to StaticAttentionmask #8736

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
5 changes: 3 additions & 2 deletions examples/models/llama/static_attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,18 +125,19 @@ def update(


class StaticAttentionMask:
def __init__(self, input_len, cache_len, style):
def __init__(self, input_len, cache_len, style, mask_val=float("-inf")):
self.input_len = input_len
self.cache_len = cache_len
assert style in ("shift_pointer", "smart_mask")
self.style = style
self.mask_val = mask_val
self.unmasked_len = 0
self.tensor = torch.zeros(1, input_len, input_len + cache_len)
self.reset()

def reset(self):
self.unmasked_len = 0
self.tensor[:, :, : self.cache_len] = float("-inf")
self.tensor[:, :, : self.cache_len] = self.mask_val

def unmask(self, new_unmasked_len):
if new_unmasked_len <= 0:
Expand Down
Loading