spaCy/spacy/train.py

84 lines
3.1 KiB
Python
Raw Normal View History

from __future__ import absolute_import
from __future__ import unicode_literals
import random
2017-03-11 20:12:48 +03:00
import tqdm
from .gold import GoldParse
from .scorer import Scorer
2016-10-13 04:24:53 +03:00
from .gold import merge_sents
class Trainer(object):
2016-10-12 15:26:02 +03:00
'''Manage training of an NLP pipeline.'''
def __init__(self, nlp, gold_tuples):
self.nlp = nlp
self.gold_tuples = gold_tuples
2017-03-11 20:12:48 +03:00
self.nr_epoch = 0
2017-03-14 23:28:43 +03:00
self.nr_itn = 0
2016-10-13 04:24:53 +03:00
def epochs(self, nr_epoch, augment_data=None, gold_preproc=False):
2016-11-25 17:55:33 +03:00
cached_golds = {}
def _epoch(indices):
2017-03-11 20:12:48 +03:00
for i in tqdm.tqdm(indices):
2016-11-25 17:55:33 +03:00
raw_text, paragraph_tuples = self.gold_tuples[i]
2016-10-13 04:24:53 +03:00
if gold_preproc:
raw_text = None
else:
paragraph_tuples = merge_sents(paragraph_tuples)
2016-11-25 17:55:33 +03:00
if augment_data is None:
docs = self.make_docs(raw_text, paragraph_tuples)
if i in cached_golds:
golds = cached_golds[i]
else:
golds = self.make_golds(docs, paragraph_tuples)
else:
raw_text, paragraph_tuples = augment_data(raw_text, paragraph_tuples)
2016-11-25 17:55:33 +03:00
docs = self.make_docs(raw_text, paragraph_tuples)
golds = self.make_golds(docs, paragraph_tuples)
for doc, gold in zip(docs, golds):
yield doc, gold
2017-03-14 23:28:43 +03:00
self.nr_itn += 1
2016-11-25 17:55:33 +03:00
indices = list(range(len(self.gold_tuples)))
for itn in range(nr_epoch):
2016-11-25 17:55:33 +03:00
random.shuffle(indices)
yield _epoch(indices)
2017-03-11 20:12:48 +03:00
self.nr_epoch += 1
def update(self, doc, gold):
2016-10-16 00:53:46 +03:00
for process in self.nlp.pipeline:
if hasattr(process, 'update'):
2017-03-14 23:28:43 +03:00
loss = process.update(doc, gold, itn=self.nr_itn)
process(doc)
return doc
2016-10-13 04:24:53 +03:00
def evaluate(self, dev_sents, gold_preproc=False):
scorer = Scorer()
for raw_text, paragraph_tuples in dev_sents:
2016-10-13 04:24:53 +03:00
if gold_preproc:
raw_text = None
else:
paragraph_tuples = merge_sents(paragraph_tuples)
docs = self.make_docs(raw_text, paragraph_tuples)
golds = self.make_golds(docs, paragraph_tuples)
for doc, gold in zip(docs, golds):
2016-11-25 17:55:33 +03:00
for process in self.nlp.pipeline:
process(doc)
scorer.score(doc, gold)
return scorer
def make_docs(self, raw_text, paragraph_tuples):
if raw_text is not None:
return [self.nlp.tokenizer(raw_text)]
else:
return [self.nlp.tokenizer.tokens_from_list(sent_tuples[0][1])
for sent_tuples in paragraph_tuples]
def make_golds(self, docs, paragraph_tuples):
if len(docs) == 1:
2016-11-25 17:55:33 +03:00
return [GoldParse.from_annot_tuples(docs[0], sent_tuples[0])
for sent_tuples in paragraph_tuples]
else:
2016-11-25 17:55:33 +03:00
return [GoldParse.from_annot_tuples(doc, sent_tuples[0])
for doc, sent_tuples in zip(docs, paragraph_tuples)]