Skip to content

Fix KTO compute_kl to average KL across all batches #216

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
EphraiemSarabamoun wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix KTO compute_kl to average KL across all batches #216

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
7 changes: 6 additions & 1 deletion align_anything/trainers/text_to_text/kto.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,8 @@ def compute_kl(self):
sampler=DistributedSampler(random_dataset, shuffle=True),
batch_size=self.cfgs.train_cfgs.per_device_kl_batch_size,
)
kl_sum = 0
num_batches = 0
for batch in self.random_dataloader:
log_probs = self.compute_log_probs( # size = (2 * B, L - 1)
self.model.module,
Expand All @@ -78,7 +80,10 @@ def compute_kl(self):
)
kl = (log_probs - ref_log_probs).mean()

self.kl = max(kl, 0)
kl_sum = kl_sum + max(kl, 0)
num_batches += 1

self.kl = kl_sum / num_batches

def loss( # pylint: disable=too-many-locals
self,
Expand Down