explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-25 17:36:30 +03:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'develop' of https://github.com/explosion/spaCy into develop

Browse Source
This commit is contained in:
ines 2017-11-01 16:49:44 +01:00
commit 1d1f91a041
3 changed files with 51 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
examples/training/train_textcat.py
View File

@ -26,8 +26,9 @@ from spacy.pipeline import TextCategorizer
@plac.annotations(
model=("Model name. Defaults to blank 'en' model.", "option", "m", str),
output_dir=("Optional output directory", "option", "o", Path),
n_examples=("Number of texts to train from", "option", "N", int),
n_iter=("Number of training iterations", "option", "n", int))
def main(model=None, output_dir=None, n_iter=20):
def main(model=None, output_dir=None, n_iter=20, n_texts=2000):
if model is not None:
nlp = spacy.load(model) # load existing spaCy model
print("Loaded model '%s'" % model)
@ -50,7 +51,8 @@ def main(model=None, output_dir=None, n_iter=20):
# load the IMBD dataset
print("Loading IMDB data...")
(train_texts, train_cats), (dev_texts, dev_cats) = load_data(limit=2000)
print("Using %d training examples" % n_texts)
(train_texts, train_cats), (dev_texts, dev_cats) = load_data(limit=n_texts)
train_docs = [nlp.tokenizer(text) for text in train_texts]
train_gold = [GoldParse(doc, cats=cats) for doc, cats in
zip(train_docs, train_cats)]
@ -65,14 +67,14 @@ def main(model=None, output_dir=None, n_iter=20):
for i in range(n_iter):
losses = {}
# batch up the examples using spaCy's minibatch
batches = minibatch(train_data, size=compounding(4., 128., 1.001))
batches = minibatch(train_data, size=compounding(4., 32., 1.001))
for batch in batches:
docs, golds = zip(*batch)
nlp.update(docs, golds, sgd=optimizer, drop=0.2, losses=losses)
with textcat.model.use_params(optimizer.averages):
# evaluate on the dev data split off in load_data()
scores = evaluate(nlp.tokenizer, textcat, dev_texts, dev_cats)
print('{0:.3f}\t{0:.3f}\t{0:.3f}\t{0:.3f}' # print a simple table
print('{0:.3f}\t{1:.3f}\t{2:.3f}\t{3:.3f}' # print a simple table
.format(losses['textcat'], scores['textcat_p'],
scores['textcat_r'], scores['textcat_f']))

2
spacy/_ml.py
View File

@ -434,7 +434,7 @@ def build_text_classifier(nr_class, width=64, **cfg):
pretrained_dims = cfg.get('pretrained_dims', 0)
with Model.define_operators({'>>': chain, '+': add, '|': concatenate,
'**': clone}):
if cfg.get('low_data'):
if cfg.get('low_data') and pretrained_dims:
model = (
SpacyVectors
>> flatten_add_lengths

48
spacy/pipeline.pyx
View File

@ -11,9 +11,9 @@ import ujson
import msgpack
from thinc.api import chain
from thinc.v2v import Softmax
from thinc.v2v import Affine, Softmax
from thinc.t2v import Pooling, max_pool, mean_pool
from thinc.neural.util import to_categorical
from thinc.neural.util import to_categorical, copy_array
from thinc.neural._classes.difference import Siamese, CauchySimilarity
from .tokens.doc cimport Doc
@ -130,6 +130,15 @@ class Pipe(object):
documents and their predicted scores."""
raise NotImplementedError
def add_label(self, label):
"""Add an output label, to be predicted by the model.
It's possible to extend pre-trained models with new labels,
but care should be taken to avoid the "catastrophic forgetting"
problem.
"""
raise NotImplementedError
def begin_training(self, gold_tuples=tuple(), pipeline=None):
"""Initialize the pipe for training, using data exampes if available.
If no model has been initialized yet, the model is added."""
@ -325,6 +334,14 @@ class Tagger(Pipe):
self.cfg.setdefault('pretrained_dims',
self.vocab.vectors.data.shape[1])
@property
def labels(self):
return self.cfg.setdefault('tag_names', [])
@labels.setter
def labels(self, value):
self.cfg['tag_names'] = value
def __call__(self, doc):
tags = self.predict([doc])
self.set_annotations([doc], tags)
@ -352,6 +369,7 @@ class Tagger(Pipe):
cdef Doc doc
cdef int idx = 0
cdef Vocab vocab = self.vocab
tags = list(self.labels)
for i, doc in enumerate(docs):
doc_tag_ids = batch_tag_ids[i]
if hasattr(doc_tag_ids, 'get'):
@ -359,7 +377,7 @@ class Tagger(Pipe):
for j, tag_id in enumerate(doc_tag_ids):
# Don't clobber preset POS tags
if doc.c[j].tag == 0 and doc.c[j].pos == 0:
vocab.morphology.assign_tag_id(&doc.c[j], tag_id)
vocab.morphology.assign_tag(&doc.c[j], tags[tag_id])
idx += 1
doc.is_tagged = True
@ -420,6 +438,17 @@ class Tagger(Pipe):
def Model(cls, n_tags, **cfg):
return build_tagger_model(n_tags, **cfg)
def add_label(self, label):
if label in self.labels:
return 0
smaller = self.model[-1]._layers[-1]
larger = Softmax(len(self.labels)+1, smaller.nI)
copy_array(larger.W[:smaller.nO], smaller.W)
copy_array(larger.b[:smaller.nO], smaller.b)
self.model[-1]._layers[-1] = larger
self.labels.append(label)
return 1
def use_params(self, params):
with self.model.use_params(params):
yield
@ -675,7 +704,7 @@ class TextCategorizer(Pipe):
@property
def labels(self):
return self.cfg.get('labels', ['LABEL'])
return self.cfg.setdefault('labels', ['LABEL'])
@labels.setter
def labels(self, value):
@ -727,6 +756,17 @@ class TextCategorizer(Pipe):
mean_square_error = ((scores-truths)**2).sum(axis=1).mean()
return mean_square_error, d_scores
def add_label(self, label):
if label in self.labels:
return 0
smaller = self.model[-1]._layers[-1]
larger = Affine(len(self.labels)+1, smaller.nI)
copy_array(larger.W[:smaller.nO], smaller.W)
copy_array(larger.b[:smaller.nO], smaller.b)
self.model[-1]._layers[-1] = larger
self.labels.append(label)
return 1
def begin_training(self, gold_tuples=tuple(), pipeline=None):
if pipeline and getattr(pipeline[0], 'name', None) == 'tensorizer':
token_vector_width = pipeline[0].model.nO