explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-26 01:46:28 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix drop_layer calculation

Browse Source
This commit is contained in:
Matthew Honnibal 2017-09-04 09:26:38 -05:00
parent 789e1a3980
commit 9f512e657a
1 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
spacy/_ml.py
View File

@ -212,12 +212,14 @@ class PrecomputableMaxouts(Model):
def drop_layer(layer, factor=2.): def drop_layer(layer, factor=2.):
def drop_layer_fwd(X, drop=0.): def drop_layer_fwd(X, drop=0.):
drop *= factor if drop <= 0.:
mask = layer.ops.get_dropout_mask((1,), drop)
if mask is None or mask > 0:
return layer.begin_update(X, drop=drop) return layer.begin_update(X, drop=drop)
else: else:
return X, lambda dX, sgd=None: dX coinflip = layer.ops.xp.random.random()
if (coinflip / factor) >= drop:
return layer.begin_update(X, drop=drop)
else:
return X, lambda dX, sgd=None: dX
model = wrap(drop_layer_fwd, layer) model = wrap(drop_layer_fwd, layer)
model.predict = layer model.predict = layer
@ -362,6 +364,8 @@ def get_token_vectors(tokens_attrs_vectors, drop=0.):
def backward(d_output, sgd=None): def backward(d_output, sgd=None):
return (tokens, d_output) return (tokens, d_output)
return vectors, backward return vectors, backward
def fine_tune(embedding, combine=None): def fine_tune(embedding, combine=None):
if combine is not None: if combine is not None:
raise NotImplementedError( raise NotImplementedError(