Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity

training loop in proper pipe format

This commit is contained in:
svlandeg 2019-06-07 15:55:10 +02:00
parent 0486ccabfd
commit 7de1ee69b8
2 changed files with 49 additions and 48 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
examples/pipeline/wiki_entity_linking/wiki_nel_pipeline.py
View File

@ -126,7 +126,7 @@ if __name__ == "__main__":
id_to_descr=id_to_descr, id_to_descr=id_to_descr,
doc_cutoff=DOC_CHAR_CUTOFF, doc_cutoff=DOC_CHAR_CUTOFF,
dev=False, dev=False,
limit=10, limit=100,
to_print=False) to_print=False)
el_pipe = nlp.create_pipe(name='entity_linker', config={"kb": my_kb}) el_pipe = nlp.create_pipe(name='entity_linker', config={"kb": my_kb})
@ -137,6 +137,8 @@ if __name__ == "__main__":
nlp.begin_training() nlp.begin_training()
for itn in range(EPOCHS): for itn in range(EPOCHS):
print()
print("EPOCH", itn)
random.shuffle(train_data) random.shuffle(train_data)
losses = {} losses = {}
batches = minibatch(train_data, size=compounding(4.0, 32.0, 1.001)) batches = minibatch(train_data, size=compounding(4.0, 32.0, 1.001))
@ -150,15 +152,6 @@ if __name__ == "__main__":
) )
print("Losses", losses) print("Losses", losses)
### BELOW CODE IS DEPRECATED ###
# STEP 6: apply the EL algorithm on the training dataset - TODO deprecated - code moved to pipes.pyx
if run_el_training:
print("STEP 6: training", datetime.datetime.now())
trainer = EL_Model(kb=my_kb, nlp=nlp)
trainer.train_model(training_dir=TRAINING_DIR, entity_descr_output=ENTITY_DESCR, trainlimit=10000, devlimit=500)
print()
# STEP 7: apply the EL algorithm on the dev dataset (TODO: overlaps with code from run_el_training ?) # STEP 7: apply the EL algorithm on the dev dataset (TODO: overlaps with code from run_el_training ?)
if apply_to_dev: if apply_to_dev:
run_el.run_el_dev(kb=my_kb, nlp=nlp, training_dir=TRAINING_DIR, limit=2000) run_el.run_el_dev(kb=my_kb, nlp=nlp, training_dir=TRAINING_DIR, limit=2000)

84
spacy/pipeline/pipes.pyx
View File

@ -1125,51 +1125,59 @@ class EntityLinker(Pipe):
docs = [docs] docs = [docs]
golds = [golds] golds = [golds]
article_docs = list()
sentence_docs = list()
entity_encodings = list()
for doc, gold in zip(docs, golds): for doc, gold in zip(docs, golds):
print("doc", doc)
for entity in gold.links: for entity in gold.links:
start, end, gold_kb = entity start, end, gold_kb = entity
print("entity", entity) mention = doc[start:end]
mention = doc[start:end].text sentence = mention.sent
print("mention", mention)
candidates = self.kb.get_candidates(mention) candidates = self.kb.get_candidates(mention.text)
for c in candidates: for c in candidates:
prior_prob = c.prior_prob
kb_id = c.entity_ kb_id = c.entity_
print("candidate", kb_id, prior_prob) # TODO: currently only training on the positive instances
entity_encoding = c.entity_vector if kb_id == gold_kb:
print() prior_prob = c.prior_prob
entity_encoding = c.entity_vector
print() entity_encodings.append(entity_encoding)
article_docs.append(doc)
sentence_docs.append(sentence.as_doc())
# entity_encodings = None #TODO if len(entity_encodings) > 0:
# doc_encodings, bp_doc = self.article_encoder.begin_update(article_docs, drop=drop) doc_encodings, bp_doc = self.article_encoder.begin_update(article_docs, drop=drop)
# sent_encodings, bp_sent = self.sent_encoder.begin_update(sentence_docs, drop=drop) sent_encodings, bp_sent = self.sent_encoder.begin_update(sentence_docs, drop=drop)
#
# concat_encodings = [list(doc_encodings[i]) + list(sent_encodings[i]) for i in concat_encodings = [list(doc_encodings[i]) + list(sent_encodings[i]) for i in
# range(len(article_docs))] range(len(article_docs))]
# mention_encodings, bp_cont = self.mention_encoder.begin_update(np.asarray(concat_encodings), drop=self.DROP) mention_encodings, bp_mention = self.mention_encoder.begin_update(np.asarray(concat_encodings), drop=drop)
#
# loss, d_scores = self.get_loss(scores=mention_encodings, golds=entity_encodings, docs=None) entity_encodings = np.asarray(entity_encodings, dtype=np.float32)
#
# mention_gradient = bp_cont(d_scores, sgd=self.sgd_cont) loss, d_scores = self.get_loss(scores=mention_encodings, golds=entity_encodings, docs=None)
#
# # gradient : concat (doc+sent) vs. desc mention_gradient = bp_mention(d_scores, sgd=self.sgd_mention)
# sent_start = self.article_encoder.nO
# sent_gradients = list() # gradient : concat (doc+sent) vs. desc
# doc_gradients = list() sent_start = self.article_encoder.nO
# for x in mention_gradient: sent_gradients = list()
# doc_gradients.append(list(x[0:sent_start])) doc_gradients = list()
# sent_gradients.append(list(x[sent_start:])) for x in mention_gradient:
# doc_gradients.append(list(x[0:sent_start]))
# bp_doc(doc_gradients, sgd=self.sgd_article) sent_gradients.append(list(x[sent_start:]))
# bp_sent(sent_gradients, sgd=self.sgd_sent)
# bp_doc(doc_gradients, sgd=self.sgd_article)
# if losses is not None: bp_sent(sent_gradients, sgd=self.sgd_sent)
# losses.setdefault(self.name, 0.0)
# losses[self.name] += loss if losses is not None:
# return loss losses.setdefault(self.name, 0.0)
return None losses[self.name] += loss
return loss
return 0
def get_loss(self, docs, golds, scores): def get_loss(self, docs, golds, scores):
loss, gradients = get_cossim_loss(scores, golds) loss, gradients = get_cossim_loss(scores, golds)