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

Avoid to loop through the whole counter in bleu_score method #1913

Merged
merged 3 commits into from
Sep 27, 2022
Merged

Avoid to loop through the whole counter in bleu_score method #1913

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6373397
avoid to loop through the whole counter in bleu_score method
Asugawara Sep 25, 2022
ad3903c
fix bug when max_n > len(candidate)
Asugawara Sep 26, 2022
e306408
add comment to explain L88
Asugawara Sep 27, 2022
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
14 changes: 8 additions & 6 deletions torchtext/data/metrics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,11 +65,12 @@ def bleu_score(candidate_corpus, references_corpus, max_n=4, weights=[0.25] * 4)
refs_len = 0.0

for (candidate, refs) in zip(candidate_corpus, references_corpus):
candidate_len += len(candidate)
current_candidate_len = len(candidate)
candidate_len += current_candidate_len

# Get the length of the reference that's closest in length to the candidate
refs_len_list = [float(len(ref)) for ref in refs]
refs_len += min(refs_len_list, key=lambda x: abs(len(candidate) - x))
refs_len += min(refs_len_list, key=lambda x: abs(current_candidate_len - x))

reference_counters = _compute_ngram_counter(refs[0], max_n)
for ref in refs[1:]:
Expand All @@ -79,11 +80,12 @@ def bleu_score(candidate_corpus, references_corpus, max_n=4, weights=[0.25] * 4)

clipped_counter = candidate_counter & reference_counters

for ngram in clipped_counter:
clipped_counts[len(ngram) - 1] += clipped_counter[ngram]
for ngram, count in clipped_counter.items():
clipped_counts[len(ngram) - 1] += count

for ngram in candidate_counter: # TODO: no need to loop through the whole counter
total_counts[len(ngram) - 1] += candidate_counter[ngram]
for i in range(max_n):
Copy link
Contributor

Choose a reason for hiding this comment

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

Can you add a one-line comment explaining the change?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Thank you for your review! Sure!

# The number of N-grams in a `candidate` of T tokens is `T - (N - 1)`
total_counts[i] += max(current_candidate_len - i, 0)

if min(clipped_counts) == 0:
return 0.0
Expand Down