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

Jump to bottom

Reduce VINC reporter file size by >1000x #219

Merged
merged 6 commits into from
May 3, 2023
Merged

Reduce VINC reporter file size by >1000x #219

Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
c79f47d
reduced reporter filesize by 4x; still unsure why the pickle file sto…
AlexTMallen Apr 27, 2023
dca33b3
add save_reporter_stats CLA
AlexTMallen Apr 27, 2023
5d49eb3
VINC reporters are now >1000x smaller on disk
norabelrose May 2, 2023
7c29973
Merge remote-tracking branch 'origin/main' into light-reporter-saving
norabelrose May 2, 2023
bdd00cd
Temporarily disable Platt scaling
norabelrose May 3, 2023
e4650cf
Merge branch 'main' into light-reporter-saving
norabelrose May 3, 2023
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
Prev Previous commit
Next Next commit
Temporarily disable Platt scaling
  • Loading branch information
@norabelrose
norabelrose committed May 3, 2023
commit bdd00cd0c86cd51618b4e45e5b62b53381a960e1
11 changes: 6 additions & 5 deletions elk/training/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,11 +82,12 @@ def apply_to_layer(
val_pair=(val_x0, val_x1),
)

(_, v, k, _) = first_train_h.shape
reporter.platt_scale(
to_one_hot(repeat(train_gt, "n -> (n v)", v=v), k).flatten(),
rearrange(first_train_h, "n v k d -> (n v k) d"),
)
# TODO: Enable Platt scaling for CCS once normalization is fixed
# (_, v, k, _) = first_train_h.shape
# reporter.platt_scale(
# to_one_hot(repeat(train_gt, "n -> (n v)", v=v), k).flatten(),
# rearrange(first_train_h, "n v k d -> (n v k) d"),
# )

elif isinstance(self.net, EigenReporterConfig):
# We set num_classes to None to enable training on datasets with different
Expand Down