explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-28 02:46:35 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix set_annotations during Parser.update

Browse Source
This commit is contained in:
Matthew Honnibal 2021-01-25 00:52:00 +11:00
parent 492c948937
commit 827fb51e6c
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
spacy/pipeline/transition_parser.pyx
View File

@ -370,7 +370,11 @@ cdef class Parser(TrainablePipe):
if sgd not in (None, False): if sgd not in (None, False):
self.finish_update(sgd) self.finish_update(sgd)
docs = [eg.predicted for eg in examples] docs = [eg.predicted for eg in examples]
self.set_annotations(docs, all_states) # TODO: Refactor so we don't have to parse twice like this (ugh)
# The issue is that we cut up the gold batch into sub-states, and that
# makes it hard to get the actual predicted transition sequence.
predicted_states = self.predict(docs)
self.set_annotations(docs, predicted_states)
# Ugh, this is annoying. If we're working on GPU, we want to free the # Ugh, this is annoying. If we're working on GPU, we want to free the
# memory ASAP. It seems that Python doesn't necessarily get around to # memory ASAP. It seems that Python doesn't necessarily get around to
# removing these in time if we don't explicitly delete? It's confusing. # removing these in time if we don't explicitly delete? It's confusing.