explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-07-18 20:22:25 +03:00
Code Issues Packages Projects Releases Wiki Activity

Change mention limit to match reference implementations

Browse Source 
This generall means fewer spans are considered, which makes individual
steps in training faster but can make training take longer to find the
good spans.
This commit is contained in:
Paul O'Leary McCann 2021-08-08 19:55:52 +09:00
parent 1d1679d431
commit 56803d3909
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
spacy/ml/models/coref.py
View File

@ -237,8 +237,8 @@ def coarse_prune(
# calculate the doc length # calculate the doc length
doclen = ends[-1] - starts[0] doclen = ends[-1] - starts[0]
# XXX seems to make more sense to use menlen than doclen here? # XXX seems to make more sense to use menlen than doclen here?
# mlimit = min(mention_limit, int(mention_limit_ratio * doclen)) # coref-hoi uses doclen (number of words).
mlimit = min(mention_limit, int(mention_limit_ratio * menlen)) mlimit = min(mention_limit, int(mention_limit_ratio * doclen))
# csel is a 1d integer list # csel is a 1d integer list
csel = select_non_crossing_spans(tops, starts, ends, mlimit) csel = select_non_crossing_spans(tops, starts, ends, mlimit)
# add the offset so these indices are absolute # add the offset so these indices are absolute