Skip to content

fix: random sampling in ForgetRetainDataset #145

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 6 commits into
base: main
Choose a base branch
from
Open

fix: random sampling in ForgetRetainDataset #145

Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
faad8d9
fix: random sampling in ForgetRetainDataset
Sep 27, 2025
c079574
feat: add seed parameter for reproducibility in ForgetRetainDataset
Oct 1, 2025
60e099a
refactor: fix lint
Oct 8, 2025
7a8b5fd
fix: ensure unique random seed per item in ForgetRetainDataset
Oct 8, 2025
19a6141
fix: remove seed arg from data pipeline and make rank-aware seeding i…
Oct 19, 2025
80ef3c7
fix: use rank-specific seeding for ForgetRetainDataset
Oct 19, 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
8 changes: 6 additions & 2 deletions src/data/unlearn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,14 +33,18 @@ def __len__(self):

def __getitem__(self, idx):
item = {}
g = torch.Generator()
rank = torch.distributed.get_rank() if torch.distributed.is_initialized() else 0
seed = int(torch.empty((), dtype=torch.int64).random_().item() + rank)
g.manual_seed(seed)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it would be better to use the seed from the experiment config here, rather than
int(torch.empty((), dtype=torch.int64).random_().item() to avoid introducing randomness uncontrolled by the seed.

can you try to see if you can make the experiment's cfg.seed available to this dataset class and then use seed = exp_seed + rank here?

if self.anchor == "forget":
item["forget"] = self.forget[idx]
if self.retain:
retain_idx = torch.randint(0, len(self.retain), (1,)).item()
retain_idx = torch.randint(0, len(self.retain), (1,), generator=g).item()
item["retain"] = self.retain[retain_idx]
elif self.anchor == "retain":
item["retain"] = self.retain[idx]
if self.forget:
forget_idx = torch.randint(0, len(self.forget), (1,)).item()
forget_idx = torch.randint(0, len(self.forget), (1,), generator=g).item()
item["forget"] = self.forget[forget_idx]
return item