explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-02-04 13:40:34 +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 2017-11-03 11:29:36 +01:00
commit 7986ada1cb
3 changed files with 25 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
spacy/_ml.py
View File

@ -150,10 +150,10 @@ class PrecomputableAffine(Model):
def _backprop_padding(self, dY, ids): def _backprop_padding(self, dY, ids):
# (1, nF, nO, nP) += (nN, nF, nO, nP) where IDs (nN, nF) < 0 # (1, nF, nO, nP) += (nN, nF, nO, nP) where IDs (nN, nF) < 0
for i in range(ids.shape[0]): mask = ids < 0.
for j in range(ids.shape[1]): mask = mask.sum(axis=1)
if ids[i,j] < 0: d_pad = dY * mask.reshape((ids.shape[0], 1, 1))
self.d_pad[0,j] += dY[i, j] self.d_pad += d_pad.sum(axis=0)
return dY, ids return dY, ids
@staticmethod @staticmethod

4
spacy/cli/train.py
View File

@ -85,6 +85,7 @@ def train(cmd, lang, output_dir, train_data, dev_data, n_iter=30, n_sents=0,
batch_sizes = util.compounding(util.env_opt('batch_from', 1), batch_sizes = util.compounding(util.env_opt('batch_from', 1),
util.env_opt('batch_to', 16), util.env_opt('batch_to', 16),
util.env_opt('batch_compound', 1.001)) util.env_opt('batch_compound', 1.001))
max_doc_len = util.env_opt('max_doc_len', 5000)
corpus = GoldCorpus(train_path, dev_path, limit=n_sents) corpus = GoldCorpus(train_path, dev_path, limit=n_sents)
n_train_words = corpus.count_train() n_train_words = corpus.count_train()
@ -108,6 +109,9 @@ def train(cmd, lang, output_dir, train_data, dev_data, n_iter=30, n_sents=0,
with tqdm.tqdm(total=n_train_words, leave=False) as pbar: with tqdm.tqdm(total=n_train_words, leave=False) as pbar:
losses = {} losses = {}
for batch in minibatch(train_docs, size=batch_sizes): for batch in minibatch(train_docs, size=batch_sizes):
batch = [(d, g) for (d, g) in batch if len(d) < max_doc_len]
if not batch:
continue
docs, golds = zip(*batch) docs, golds = zip(*batch)
nlp.update(docs, golds, sgd=optimizer, nlp.update(docs, golds, sgd=optimizer,
drop=next(dropout_rates), losses=losses) drop=next(dropout_rates), losses=losses)

17
spacy/tests/test_misc.py
View File

@ -6,6 +6,7 @@ from .. import util
from ..displacy import parse_deps, parse_ents from ..displacy import parse_deps, parse_ents
from ..tokens import Span from ..tokens import Span
from .util import get_doc from .util import get_doc
from .._ml import PrecomputableAffine
from pathlib import Path from pathlib import Path
import pytest import pytest
@ -59,3 +60,19 @@ def test_displacy_parse_deps(en_vocab):
assert deps['arcs'] == [{'start': 0, 'end': 1, 'label': 'nsubj', 'dir': 'left'}, assert deps['arcs'] == [{'start': 0, 'end': 1, 'label': 'nsubj', 'dir': 'left'},
{'start': 2, 'end': 3, 'label': 'det', 'dir': 'left'}, {'start': 2, 'end': 3, 'label': 'det', 'dir': 'left'},
{'start': 1, 'end': 3, 'label': 'attr', 'dir': 'right'}] {'start': 1, 'end': 3, 'label': 'attr', 'dir': 'right'}]
def test_PrecomputableAffine(nO=4, nI=5, nF=3, nP=2):
model = PrecomputableAffine(nO=nO, nI=nI, nF=nF, nP=nP)
assert model.W.shape == (nF, nO, nP, nI)
tensor = model.ops.allocate((10, nI))
Y, get_dX = model.begin_update(tensor)
assert Y.shape == (tensor.shape[0]+1, nF, nO, nP)
assert model.d_pad.shape == (1, nF, nO, nP)
dY = model.ops.allocate((15, nF, nO, nP))
ids = model.ops.allocate((15, nF))
ids[1,2] = -1
dY[1,2] = 1
assert model.d_pad[0, 2, 0, 0] == 0.
model._backprop_padding(dY, ids)
assert model.d_pad[0, 2, 0, 0] == 1.