explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-02-28 17:40:39 +03:00
Code Issues Packages Projects Releases Wiki Activity

Restore beam_density argument for parser beam

Browse Source
This commit is contained in:
Matthew Honnibal 2018-05-15 17:54:39 +02:00
parent 625ee6c464
commit 7b9195657b
3 changed files with 24 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
spacy/syntax/_beam_utils.pyx
View File

@ -68,7 +68,7 @@ cdef class ParserBeam(object):
cdef StateClass state cdef StateClass state
cdef StateC* st cdef StateC* st
for state in states: for state in states:
beam = Beam(self.moves.n_moves, width, density) beam = Beam(self.moves.n_moves, width, min_density=density)
beam.initialize(self.moves.init_beam_state, state.c.length, beam.initialize(self.moves.init_beam_state, state.c.length,
state.c._sent) state.c._sent)
for i in range(beam.width): for i in range(beam.width):
@ -161,12 +161,12 @@ def update_beam(TransitionSystem moves, int nr_feature, int max_steps,
states, golds, states, golds,
state2vec, vec2scores, state2vec, vec2scores,
int width, losses=None, drop=0., int width, losses=None, drop=0.,
early_update=True): early_update=True, beam_density=0.0):
global nr_update global nr_update
cdef MaxViolation violn cdef MaxViolation violn
nr_update += 1 nr_update += 1
pbeam = ParserBeam(moves, states, golds, width=width) pbeam = ParserBeam(moves, states, golds, width=width, density=beam_density)
gbeam = ParserBeam(moves, states, golds, width=width) gbeam = ParserBeam(moves, states, golds, width=width, density=beam_density)
cdef StateClass state cdef StateClass state
beam_maps = [] beam_maps = []
backprops = [] backprops = []

29
spacy/syntax/nn_parser.pyx
View File

@ -182,7 +182,9 @@ cdef class Parser:
""" """
if beam_width is None: if beam_width is None:
beam_width = self.cfg.get('beam_width', 1) beam_width = self.cfg.get('beam_width', 1)
states = self.predict([doc], beam_width=beam_width) beam_density = self.cfg.get('beam_density', 0.)
states = self.predict([doc], beam_width=beam_width,
beam_density=beam_density)
self.set_annotations([doc], states, tensors=None) self.set_annotations([doc], states, tensors=None)
return doc return doc
@ -197,24 +199,27 @@ cdef class Parser:
""" """
if beam_width is None: if beam_width is None:
beam_width = self.cfg.get('beam_width', 1) beam_width = self.cfg.get('beam_width', 1)
beam_density = self.cfg.get('beam_density', 0.)
cdef Doc doc cdef Doc doc
for batch in cytoolz.partition_all(batch_size, docs): for batch in cytoolz.partition_all(batch_size, docs):
batch_in_order = list(batch) batch_in_order = list(batch)
by_length = sorted(batch_in_order, key=lambda doc: len(doc)) by_length = sorted(batch_in_order, key=lambda doc: len(doc))
for subbatch in cytoolz.partition_all(8, by_length): for subbatch in cytoolz.partition_all(8, by_length):
subbatch = list(subbatch) subbatch = list(subbatch)
parse_states = self.predict(subbatch, beam_width=beam_width) parse_states = self.predict(subbatch, beam_width=beam_width,
beam_density=beam_density)
self.set_annotations(subbatch, parse_states, tensors=None) self.set_annotations(subbatch, parse_states, tensors=None)
for doc in batch_in_order: for doc in batch_in_order:
yield doc yield doc
def predict(self, docs, beam_width=1, drop=0.): def predict(self, docs, beam_width=1, beam_density=0.0, drop=0.):
if isinstance(docs, Doc): if isinstance(docs, Doc):
docs = [docs] docs = [docs]
if beam_width < 2: if beam_width < 2:
return self.greedy_parse(docs, drop=drop) return self.greedy_parse(docs, drop=drop)
else: else:
return self.beam_parse(docs, beam_width=beam_width, drop=drop) return self.beam_parse(docs, beam_width=beam_width,
beam_density=beam_density, drop=drop)
def greedy_parse(self, docs, drop=0.): def greedy_parse(self, docs, drop=0.):
cdef vector[StateC*] states cdef vector[StateC*] states
@ -231,12 +236,12 @@ cdef class Parser:
weights, sizes) weights, sizes)
return batch return batch
def beam_parse(self, docs, int beam_width, float drop=0.): def beam_parse(self, docs, int beam_width, float drop=0., beam_density=0.):
cdef Beam beam cdef Beam beam
cdef Doc doc cdef Doc doc
cdef np.ndarray token_ids cdef np.ndarray token_ids
model = self.model(docs) model = self.model(docs)
beams = self.moves.init_beams(docs, beam_width) beams = self.moves.init_beams(docs, beam_width, beam_density=beam_density)
token_ids = numpy.zeros((len(docs) * beam_width, self.nr_feature), token_ids = numpy.zeros((len(docs) * beam_width, self.nr_feature),
dtype='i', order='C') dtype='i', order='C')
cdef int* c_ids cdef int* c_ids
@ -358,9 +363,9 @@ cdef class Parser:
# a greedy update # a greedy update
beam_update_prob = self.cfg.get('beam_update_prob', 1.0) beam_update_prob = self.cfg.get('beam_update_prob', 1.0)
if self.cfg.get('beam_width', 1) >= 2 and numpy.random.random() < beam_update_prob: if self.cfg.get('beam_width', 1) >= 2 and numpy.random.random() < beam_update_prob:
return self.update_beam(docs, golds, return self.update_beam(docs, golds, self.cfg.get('beam_width', 1),
self.cfg['beam_width'], drop=drop, sgd=sgd, losses=losses,
drop=drop, sgd=sgd, losses=losses) beam_density=self.cfg.get('beam_density', 0.0))
# Chop sequences into lengths of this many transitions, to make the # Chop sequences into lengths of this many transitions, to make the
# batch uniform length. # batch uniform length.
cut_gold = numpy.random.choice(range(20, 100)) cut_gold = numpy.random.choice(range(20, 100))
@ -384,7 +389,8 @@ cdef class Parser:
finish_update(golds, sgd=sgd) finish_update(golds, sgd=sgd)
return losses return losses
def update_beam(self, docs, golds, width, drop=0., sgd=None, losses=None): def update_beam(self, docs, golds, width, drop=0., sgd=None, losses=None,
beam_density=0.0):
lengths = [len(d) for d in docs] lengths = [len(d) for d in docs]
states = self.moves.init_batch(docs) states = self.moves.init_batch(docs)
for gold in golds: for gold in golds:
@ -392,7 +398,8 @@ cdef class Parser:
model, finish_update = self.model.begin_update(docs, drop=drop) model, finish_update = self.model.begin_update(docs, drop=drop)
states_d_scores, backprops, beams = _beam_utils.update_beam( states_d_scores, backprops, beams = _beam_utils.update_beam(
self.moves, self.nr_feature, 10000, states, golds, model.state2vec, self.moves, self.nr_feature, 10000, states, golds, model.state2vec,
model.vec2scores, width, drop=drop, losses=losses) model.vec2scores, width, drop=drop, losses=losses,
beam_density=beam_density)
for i, d_scores in enumerate(states_d_scores): for i, d_scores in enumerate(states_d_scores):
losses[self.name] += (d_scores**2).sum() losses[self.name] += (d_scores**2).sum()
ids, bp_vectors, bp_scores = backprops[i] ids, bp_vectors, bp_scores = backprops[i]

4
spacy/syntax/transition_system.pyx
View File

@ -60,12 +60,12 @@ cdef class TransitionSystem:
offset += len(doc) offset += len(doc)
return states return states
def init_beams(self, docs, beam_width): def init_beams(self, docs, beam_width, beam_density=0.):
cdef Doc doc cdef Doc doc
beams = [] beams = []
cdef int offset = 0 cdef int offset = 0
for doc in docs: for doc in docs:
beam = Beam(self.n_moves, beam_width) beam = Beam(self.n_moves, beam_width, min_density=beam_density)
beam.initialize(self.init_beam_state, doc.length, doc.c) beam.initialize(self.init_beam_state, doc.length, doc.c)
for i in range(beam.width): for i in range(beam.width):
state = <StateC*>beam.at(i) state = <StateC*>beam.at(i)