Skip to content

Fix IndexError in LogProbTokenNorm when choices_tokens is shorter than choices_logprob #1171

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

Fix IndexError in LogProbTokenNorm when choices_tokens is shorter than choices_logprob #1171

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
16 changes: 13 additions & 3 deletions src/lighteval/metrics/normalizations.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -523,9 +523,19 @@ def normalize_log_probs(
normalized_log_probs = [choices_logprob[ix] / len(choice) for ix, choice in enumerate(choices_text)]
case LogProbTokenNorm():
assert choices_tokens is not None, "choices_tokens must be provided for token normalization"
normalized_log_probs = [
choices_logprob[ix] / len(choices_tokens[ix]) for ix in range(len(choices_logprob))
]
# Handle cases where choices_tokens might be shorter than choices_logprob
# (e.g., when token generation fails for some choices)
normalized_log_probs = []
for ix in range(len(choices_logprob)):
if ix < len(choices_tokens) and choices_tokens[ix]:
# Divide by the number of tokens (filtering out padding tokens marked as -1)
token_count = sum(1 for token in choices_tokens[ix] if token != -1)
# Avoid division by zero; use token count or 1 as fallback
token_count = max(token_count, 1)
normalized_log_probs.append(choices_logprob[ix] / token_count)
else:
# If tokens are missing, use the log probability as-is (no normalization)
normalized_log_probs.append(choices_logprob[ix])
case LogProbPMINorm():
assert unconditioned_logprob is not None, "unconditioned_logprob must be provided for PMI normalization"
normalized_log_probs = [
Expand Down