explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-11-04 09:57:26 +03:00
Code Issues Packages Projects Releases Wiki Activity

Clarify parser model CPU/GPU code

Browse Source 
The previous version worked with previous thinc, but only
because some thinc ops happened to have gpu/cpu compatible
implementations. It's better to call the right Ops instance.
This commit is contained in:
Matthw Honnibal 2019-10-20 17:15:17 +02:00
parent ee56c6a4e1
commit 3a67aa857e
1 changed files with 18 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
spacy/syntax/_parser_model.pyx
View File

@ -19,7 +19,7 @@ from thinc.extra.search cimport Beam
from thinc.api import chain, clone from thinc.api import chain, clone
from thinc.v2v import Model, Maxout, Affine from thinc.v2v import Model, Maxout, Affine
from thinc.misc import LayerNorm from thinc.misc import LayerNorm
from thinc.neural.ops import CupyOps from thinc.neural.ops import CupyOps, NumpyOps
from thinc.neural.util import get_array_module from thinc.neural.util import get_array_module
from thinc.linalg cimport Vec, VecVec from thinc.linalg cimport Vec, VecVec
cimport blis.cy cimport blis.cy
@ -425,28 +425,38 @@ cdef class precompute_hiddens:
def backward(d_state_vector_ids, sgd=None): def backward(d_state_vector_ids, sgd=None):
d_state_vector, token_ids = d_state_vector_ids d_state_vector, token_ids = d_state_vector_ids
d_state_vector = bp_nonlinearity(d_state_vector, sgd) d_state_vector = bp_nonlinearity(d_state_vector, sgd)
# This will usually be on GPU
if not isinstance(d_state_vector, self.ops.xp.ndarray):
d_state_vector = self.ops.xp.array(d_state_vector)
d_tokens = bp_hiddens((d_state_vector, token_ids), sgd) d_tokens = bp_hiddens((d_state_vector, token_ids), sgd)
return d_tokens return d_tokens
return state_vector, backward return state_vector, backward
def _nonlinearity(self, state_vector): def _nonlinearity(self, state_vector):
if isinstance(state_vector, numpy.ndarray):
ops = NumpyOps()
else:
ops = CupyOps()
if self.nP == 1: if self.nP == 1:
state_vector = state_vector.reshape(state_vector.shape[:-1]) state_vector = state_vector.reshape(state_vector.shape[:-1])
mask = state_vector >= 0. mask = state_vector >= 0.
state_vector *= mask state_vector *= mask
else: else:
state_vector, mask = self.ops.maxout(state_vector) state_vector, mask = ops.maxout(state_vector)
def backprop_nonlinearity(d_best, sgd=None): def backprop_nonlinearity(d_best, sgd=None):
if isinstance(d_best, numpy.ndarray):
ops = NumpyOps()
else:
ops = CupyOps()
mask_ = ops.asarray(mask)
# This will usually be on GPU
d_best = ops.asarray(d_best)
# Fix nans (which can occur from unseen classes.) # Fix nans (which can occur from unseen classes.)
d_best[self.ops.xp.isnan(d_best)] = 0. d_best[ops.xp.isnan(d_best)] = 0.
if self.nP == 1: if self.nP == 1:
d_best *= mask d_best *= mask_
d_best = d_best.reshape((d_best.shape + (1,))) d_best = d_best.reshape((d_best.shape + (1,)))
return d_best return d_best
else: else:
return self.ops.backprop_maxout(d_best, mask, self.nP) return ops.backprop_maxout(d_best, mask_, self.nP)
return state_vector, backprop_nonlinearity return state_vector, backprop_nonlinearity