2020-03-02 13:48:10 +03:00
|
|
|
# cython: infer_types=True, cdivision=True, boundscheck=False
|
2018-05-15 23:17:29 +03:00
|
|
|
cimport cython.parallel
|
|
|
|
cimport numpy as np
|
|
|
|
from libc.math cimport exp
|
|
|
|
from libcpp.vector cimport vector
|
|
|
|
from libc.string cimport memset, memcpy
|
|
|
|
from libc.stdlib cimport calloc, free, realloc
|
|
|
|
from cymem.cymem cimport Pool
|
|
|
|
from thinc.extra.search cimport Beam
|
2020-01-29 19:06:46 +03:00
|
|
|
from thinc.backends.linalg cimport Vec, VecVec
|
💫 Use Blis for matrix multiplications (#2966)
Our epic matrix multiplication odyssey is drawing to a close...
I've now finally got the Blis linear algebra routines in a self-contained Python package, with wheels for Windows, Linux and OSX. The only missing platform at the moment is Windows Python 2.7. The result is at https://github.com/explosion/cython-blis
Thinc v7.0.0 will make the change to Blis. I've put a Thinc v7.0.0.dev0 up on PyPi so that we can test these changes with the CI, and even get them out to spacy-nightly, before Thinc v7.0.0 is released. This PR also updates the other dependencies to be in line with the current versions master is using. I've also resolved the msgpack deprecation problems, and gotten spaCy and Thinc up to date with the latest Cython.
The point of switching to Blis is to have control of how our matrix multiplications are executed across platforms. When we were using numpy for this, a different library would be used on pip and conda, OSX would use Accelerate, etc. This would open up different bugs and performance problems, especially when multi-threading was introduced.
With the change to Blis, we now strictly single-thread the matrix multiplications. This will make it much easier to use multiprocessing to parallelise the runtime, since we won't have nested parallelism problems to deal with.
* Use blis
* Use -2 arg to Cython
* Update dependencies
* Fix requirements
* Update setup dependencies
* Fix requirement typo
* Fix msgpack errors
* Remove Python27 test from Appveyor, until Blis works there
* Auto-format setup.py
* Fix murmurhash version
2018-11-27 02:44:04 +03:00
|
|
|
cimport blis.cy
|
2018-05-15 23:17:29 +03:00
|
|
|
|
2020-03-02 13:48:10 +03:00
|
|
|
import numpy
|
|
|
|
import numpy.random
|
2020-05-18 23:23:33 +03:00
|
|
|
from thinc.api import Linear, Model, CupyOps, NumpyOps, use_ops, noop
|
2020-03-02 13:48:10 +03:00
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
from ..typedefs cimport weight_t, class_t, hash_t
|
2018-05-15 23:17:29 +03:00
|
|
|
from ..tokens.doc cimport Doc
|
|
|
|
from .stateclass cimport StateClass
|
|
|
|
from .transition_system cimport Transition
|
2020-03-02 13:48:10 +03:00
|
|
|
|
|
|
|
from ..compat import copy_array
|
|
|
|
from ..errors import Errors, TempErrors
|
|
|
|
from ..util import link_vectors_to_models, create_default_optimizer
|
|
|
|
from .. import util
|
2018-05-15 23:17:29 +03:00
|
|
|
from . import nonproj
|
|
|
|
|
|
|
|
|
|
|
|
cdef WeightsC get_c_weights(model) except *:
|
|
|
|
cdef WeightsC output
|
|
|
|
cdef precompute_hiddens state2vec = model.state2vec
|
|
|
|
output.feat_weights = state2vec.get_feat_weights()
|
|
|
|
output.feat_bias = <const float*>state2vec.bias.data
|
2019-11-19 17:54:34 +03:00
|
|
|
cdef np.ndarray vec2scores_W
|
|
|
|
cdef np.ndarray vec2scores_b
|
|
|
|
if model.vec2scores is None:
|
|
|
|
output.hidden_weights = NULL
|
|
|
|
output.hidden_bias = NULL
|
|
|
|
else:
|
2020-01-29 19:06:46 +03:00
|
|
|
vec2scores_W = model.vec2scores.get_param("W")
|
|
|
|
vec2scores_b = model.vec2scores.get_param("b")
|
2019-11-19 17:54:34 +03:00
|
|
|
output.hidden_weights = <const float*>vec2scores_W.data
|
|
|
|
output.hidden_bias = <const float*>vec2scores_b.data
|
2019-02-24 18:41:41 +03:00
|
|
|
cdef np.ndarray class_mask = model._class_mask
|
|
|
|
output.seen_classes = <const float*>class_mask.data
|
2018-05-15 23:17:29 +03:00
|
|
|
return output
|
|
|
|
|
|
|
|
|
|
|
|
cdef SizesC get_c_sizes(model, int batch_size) except *:
|
|
|
|
cdef SizesC output
|
|
|
|
output.states = batch_size
|
2019-11-19 17:54:34 +03:00
|
|
|
if model.vec2scores is None:
|
2020-01-29 19:06:46 +03:00
|
|
|
output.classes = model.state2vec.get_dim("nO")
|
2019-11-19 17:54:34 +03:00
|
|
|
else:
|
2020-01-29 19:06:46 +03:00
|
|
|
output.classes = model.vec2scores.get_dim("nO")
|
|
|
|
output.hiddens = model.state2vec.get_dim("nO")
|
|
|
|
output.pieces = model.state2vec.get_dim("nP")
|
|
|
|
output.feats = model.state2vec.get_dim("nF")
|
2018-05-15 23:17:29 +03:00
|
|
|
output.embed_width = model.tokvecs.shape[1]
|
|
|
|
return output
|
|
|
|
|
|
|
|
|
2019-10-22 16:06:44 +03:00
|
|
|
cdef ActivationsC alloc_activations(SizesC n) nogil:
|
|
|
|
cdef ActivationsC A
|
|
|
|
memset(&A, 0, sizeof(A))
|
|
|
|
resize_activations(&A, n)
|
|
|
|
return A
|
|
|
|
|
|
|
|
|
|
|
|
cdef void free_activations(const ActivationsC* A) nogil:
|
|
|
|
free(A.token_ids)
|
|
|
|
free(A.scores)
|
|
|
|
free(A.unmaxed)
|
|
|
|
free(A.hiddens)
|
|
|
|
free(A.is_valid)
|
|
|
|
|
|
|
|
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef void resize_activations(ActivationsC* A, SizesC n) nogil:
|
|
|
|
if n.states <= A._max_size:
|
|
|
|
A._curr_size = n.states
|
|
|
|
return
|
|
|
|
if A._max_size == 0:
|
|
|
|
A.token_ids = <int*>calloc(n.states * n.feats, sizeof(A.token_ids[0]))
|
|
|
|
A.scores = <float*>calloc(n.states * n.classes, sizeof(A.scores[0]))
|
|
|
|
A.unmaxed = <float*>calloc(n.states * n.hiddens * n.pieces, sizeof(A.unmaxed[0]))
|
|
|
|
A.hiddens = <float*>calloc(n.states * n.hiddens, sizeof(A.hiddens[0]))
|
|
|
|
A.is_valid = <int*>calloc(n.states * n.classes, sizeof(A.is_valid[0]))
|
|
|
|
A._max_size = n.states
|
|
|
|
else:
|
|
|
|
A.token_ids = <int*>realloc(A.token_ids,
|
|
|
|
n.states * n.feats * sizeof(A.token_ids[0]))
|
|
|
|
A.scores = <float*>realloc(A.scores,
|
|
|
|
n.states * n.classes * sizeof(A.scores[0]))
|
|
|
|
A.unmaxed = <float*>realloc(A.unmaxed,
|
|
|
|
n.states * n.hiddens * n.pieces * sizeof(A.unmaxed[0]))
|
|
|
|
A.hiddens = <float*>realloc(A.hiddens,
|
|
|
|
n.states * n.hiddens * sizeof(A.hiddens[0]))
|
|
|
|
A.is_valid = <int*>realloc(A.is_valid,
|
|
|
|
n.states * n.classes * sizeof(A.is_valid[0]))
|
|
|
|
A._max_size = n.states
|
|
|
|
A._curr_size = n.states
|
|
|
|
|
|
|
|
|
|
|
|
cdef void predict_states(ActivationsC* A, StateC** states,
|
|
|
|
const WeightsC* W, SizesC n) nogil:
|
2019-11-19 17:54:34 +03:00
|
|
|
cdef double one = 1.0
|
2018-05-15 23:17:29 +03:00
|
|
|
resize_activations(A, n)
|
|
|
|
for i in range(n.states):
|
|
|
|
states[i].set_context_tokens(&A.token_ids[i*n.feats], n.feats)
|
2019-11-19 17:54:34 +03:00
|
|
|
memset(A.unmaxed, 0, n.states * n.hiddens * n.pieces * sizeof(float))
|
|
|
|
memset(A.hiddens, 0, n.states * n.hiddens * sizeof(float))
|
2018-05-15 23:17:29 +03:00
|
|
|
sum_state_features(A.unmaxed,
|
|
|
|
W.feat_weights, A.token_ids, n.states, n.feats, n.hiddens * n.pieces)
|
|
|
|
for i in range(n.states):
|
|
|
|
VecVec.add_i(&A.unmaxed[i*n.hiddens*n.pieces],
|
|
|
|
W.feat_bias, 1., n.hiddens * n.pieces)
|
|
|
|
for j in range(n.hiddens):
|
|
|
|
index = i * n.hiddens * n.pieces + j * n.pieces
|
|
|
|
which = Vec.arg_max(&A.unmaxed[index], n.pieces)
|
|
|
|
A.hiddens[i*n.hiddens + j] = A.unmaxed[index + which]
|
|
|
|
memset(A.scores, 0, n.states * n.classes * sizeof(float))
|
2019-11-19 17:54:34 +03:00
|
|
|
if W.hidden_weights == NULL:
|
|
|
|
memcpy(A.scores, A.hiddens, n.states * n.classes * sizeof(float))
|
|
|
|
else:
|
|
|
|
# Compute hidden-to-output
|
|
|
|
blis.cy.gemm(blis.cy.NO_TRANSPOSE, blis.cy.TRANSPOSE,
|
|
|
|
n.states, n.classes, n.hiddens, one,
|
|
|
|
<float*>A.hiddens, n.hiddens, 1,
|
|
|
|
<float*>W.hidden_weights, n.hiddens, 1,
|
|
|
|
one,
|
|
|
|
<float*>A.scores, n.classes, 1)
|
|
|
|
# Add bias
|
|
|
|
for i in range(n.states):
|
|
|
|
VecVec.add_i(&A.scores[i*n.classes],
|
|
|
|
W.hidden_bias, 1., n.classes)
|
2019-02-24 18:41:41 +03:00
|
|
|
# Set unseen classes to minimum value
|
|
|
|
i = 0
|
|
|
|
min_ = A.scores[0]
|
|
|
|
for i in range(1, n.states * n.classes):
|
|
|
|
if A.scores[i] < min_:
|
|
|
|
min_ = A.scores[i]
|
|
|
|
for i in range(n.states):
|
|
|
|
for j in range(n.classes):
|
|
|
|
if not W.seen_classes[j]:
|
|
|
|
A.scores[i*n.classes+j] = min_
|
2018-05-15 23:17:29 +03:00
|
|
|
|
💫 Replace ujson, msgpack and dill/pickle/cloudpickle with srsly (#3003)
Remove hacks and wrappers, keep code in sync across our libraries and move spaCy a few steps closer to only depending on packages with binary wheels 🎉
See here: https://github.com/explosion/srsly
Serialization is hard, especially across Python versions and multiple platforms. After dealing with many subtle bugs over the years (encodings, locales, large files) our libraries like spaCy and Prodigy have steadily grown a number of utility functions to wrap the multiple serialization formats we need to support (especially json, msgpack and pickle). These wrapping functions ended up duplicated across our codebases, so we wanted to put them in one place.
At the same time, we noticed that having a lot of small dependencies was making maintainence harder, and making installation slower. To solve this, we've made srsly standalone, by including the component packages directly within it. This way we can provide all the serialization utilities we need in a single binary wheel.
srsly currently includes forks of the following packages:
ujson
msgpack
msgpack-numpy
cloudpickle
* WIP: replace json/ujson with srsly
* Replace ujson in examples
Use regular json instead of srsly to make code easier to read and follow
* Update requirements
* Fix imports
* Fix typos
* Replace msgpack with srsly
* Fix warning
2018-12-03 03:28:22 +03:00
|
|
|
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef void sum_state_features(float* output,
|
|
|
|
const float* cached, const int* token_ids, int B, int F, int O) nogil:
|
|
|
|
cdef int idx, b, f, i
|
|
|
|
cdef const float* feature
|
|
|
|
padding = cached
|
|
|
|
cached += F * O
|
|
|
|
cdef int id_stride = F*O
|
|
|
|
cdef float one = 1.
|
|
|
|
for b in range(B):
|
|
|
|
for f in range(F):
|
|
|
|
if token_ids[f] < 0:
|
|
|
|
feature = &padding[f*O]
|
|
|
|
else:
|
|
|
|
idx = token_ids[f] * id_stride + f*O
|
|
|
|
feature = &cached[idx]
|
💫 Use Blis for matrix multiplications (#2966)
Our epic matrix multiplication odyssey is drawing to a close...
I've now finally got the Blis linear algebra routines in a self-contained Python package, with wheels for Windows, Linux and OSX. The only missing platform at the moment is Windows Python 2.7. The result is at https://github.com/explosion/cython-blis
Thinc v7.0.0 will make the change to Blis. I've put a Thinc v7.0.0.dev0 up on PyPi so that we can test these changes with the CI, and even get them out to spacy-nightly, before Thinc v7.0.0 is released. This PR also updates the other dependencies to be in line with the current versions master is using. I've also resolved the msgpack deprecation problems, and gotten spaCy and Thinc up to date with the latest Cython.
The point of switching to Blis is to have control of how our matrix multiplications are executed across platforms. When we were using numpy for this, a different library would be used on pip and conda, OSX would use Accelerate, etc. This would open up different bugs and performance problems, especially when multi-threading was introduced.
With the change to Blis, we now strictly single-thread the matrix multiplications. This will make it much easier to use multiprocessing to parallelise the runtime, since we won't have nested parallelism problems to deal with.
* Use blis
* Use -2 arg to Cython
* Update dependencies
* Fix requirements
* Update setup dependencies
* Fix requirement typo
* Fix msgpack errors
* Remove Python27 test from Appveyor, until Blis works there
* Auto-format setup.py
* Fix murmurhash version
2018-11-27 02:44:04 +03:00
|
|
|
blis.cy.axpyv(blis.cy.NO_CONJUGATE, O, one,
|
|
|
|
<float*>feature, 1,
|
|
|
|
&output[b*O], 1)
|
2018-05-15 23:17:29 +03:00
|
|
|
token_ids += F
|
|
|
|
|
|
|
|
|
|
|
|
cdef void cpu_log_loss(float* d_scores,
|
|
|
|
const float* costs, const int* is_valid, const float* scores,
|
|
|
|
int O) nogil:
|
|
|
|
"""Do multi-label log loss"""
|
|
|
|
cdef double max_, gmax, Z, gZ
|
|
|
|
best = arg_max_if_gold(scores, costs, is_valid, O)
|
2019-03-15 17:22:16 +03:00
|
|
|
guess = Vec.arg_max(scores, O)
|
2019-03-10 16:53:03 +03:00
|
|
|
if best == -1 or guess == -1:
|
|
|
|
# These shouldn't happen, but if they do, we want to make sure we don't
|
|
|
|
# cause an OOB access.
|
|
|
|
return
|
2018-05-15 23:17:29 +03:00
|
|
|
Z = 1e-10
|
|
|
|
gZ = 1e-10
|
|
|
|
max_ = scores[guess]
|
|
|
|
gmax = scores[best]
|
|
|
|
for i in range(O):
|
2019-03-15 17:22:16 +03:00
|
|
|
Z += exp(scores[i] - max_)
|
|
|
|
if costs[i] <= costs[best]:
|
|
|
|
gZ += exp(scores[i] - gmax)
|
2018-05-15 23:17:29 +03:00
|
|
|
for i in range(O):
|
2019-03-15 17:22:16 +03:00
|
|
|
if costs[i] <= costs[best]:
|
2018-05-15 23:17:29 +03:00
|
|
|
d_scores[i] = (exp(scores[i]-max_) / Z) - (exp(scores[i]-gmax)/gZ)
|
|
|
|
else:
|
|
|
|
d_scores[i] = exp(scores[i]-max_) / Z
|
|
|
|
|
💫 Replace ujson, msgpack and dill/pickle/cloudpickle with srsly (#3003)
Remove hacks and wrappers, keep code in sync across our libraries and move spaCy a few steps closer to only depending on packages with binary wheels 🎉
See here: https://github.com/explosion/srsly
Serialization is hard, especially across Python versions and multiple platforms. After dealing with many subtle bugs over the years (encodings, locales, large files) our libraries like spaCy and Prodigy have steadily grown a number of utility functions to wrap the multiple serialization formats we need to support (especially json, msgpack and pickle). These wrapping functions ended up duplicated across our codebases, so we wanted to put them in one place.
At the same time, we noticed that having a lot of small dependencies was making maintainence harder, and making installation slower. To solve this, we've made srsly standalone, by including the component packages directly within it. This way we can provide all the serialization utilities we need in a single binary wheel.
srsly currently includes forks of the following packages:
ujson
msgpack
msgpack-numpy
cloudpickle
* WIP: replace json/ujson with srsly
* Replace ujson in examples
Use regular json instead of srsly to make code easier to read and follow
* Update requirements
* Fix imports
* Fix typos
* Replace msgpack with srsly
* Fix warning
2018-12-03 03:28:22 +03:00
|
|
|
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef int arg_max_if_gold(const weight_t* scores, const weight_t* costs,
|
|
|
|
const int* is_valid, int n) nogil:
|
|
|
|
# Find minimum cost
|
|
|
|
cdef float cost = 1
|
|
|
|
for i in range(n):
|
|
|
|
if is_valid[i] and costs[i] < cost:
|
|
|
|
cost = costs[i]
|
|
|
|
# Now find best-scoring with that cost
|
|
|
|
cdef int best = -1
|
|
|
|
for i in range(n):
|
|
|
|
if costs[i] <= cost and is_valid[i]:
|
|
|
|
if best == -1 or scores[i] > scores[best]:
|
|
|
|
best = i
|
|
|
|
return best
|
|
|
|
|
|
|
|
|
|
|
|
cdef int arg_max_if_valid(const weight_t* scores, const int* is_valid, int n) nogil:
|
|
|
|
cdef int best = -1
|
|
|
|
for i in range(n):
|
|
|
|
if is_valid[i] >= 1:
|
|
|
|
if best == -1 or scores[i] > scores[best]:
|
|
|
|
best = i
|
|
|
|
return best
|
|
|
|
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
|
2018-05-15 23:17:29 +03:00
|
|
|
class ParserStepModel(Model):
|
2020-05-18 23:23:33 +03:00
|
|
|
def __init__(self, docs, layers, *, has_upper, unseen_classes=None, train=True):
|
2020-01-29 19:06:46 +03:00
|
|
|
Model.__init__(self, name="parser_step_model", forward=step_forward)
|
2020-05-18 23:23:33 +03:00
|
|
|
self.attrs["has_upper"] = has_upper
|
2020-01-29 19:06:46 +03:00
|
|
|
self.tokvecs, self.bp_tokvecs = layers[0](docs, is_train=train)
|
|
|
|
if layers[1].get_dim("nP") >= 2:
|
2019-11-19 17:54:34 +03:00
|
|
|
activation = "maxout"
|
2020-05-18 23:23:33 +03:00
|
|
|
elif has_upper:
|
2019-11-19 17:54:34 +03:00
|
|
|
activation = None
|
|
|
|
else:
|
|
|
|
activation = "relu"
|
2018-05-15 23:17:29 +03:00
|
|
|
self.state2vec = precompute_hiddens(len(docs), self.tokvecs, layers[1],
|
2020-01-29 19:06:46 +03:00
|
|
|
activation=activation, train=train)
|
2020-05-18 23:23:33 +03:00
|
|
|
if has_upper:
|
2019-11-19 17:54:34 +03:00
|
|
|
self.vec2scores = layers[-1]
|
|
|
|
else:
|
|
|
|
self.vec2scores = None
|
|
|
|
self.cuda_stream = util.get_cuda_stream(non_blocking=True)
|
2018-05-15 23:17:29 +03:00
|
|
|
self.backprops = []
|
2020-05-18 23:23:33 +03:00
|
|
|
self._class_mask = numpy.zeros((self.nO,), dtype='f')
|
2019-02-24 18:41:41 +03:00
|
|
|
self._class_mask.fill(1)
|
|
|
|
if unseen_classes is not None:
|
|
|
|
for class_ in unseen_classes:
|
|
|
|
self._class_mask[class_] = 0.
|
2018-05-15 23:17:29 +03:00
|
|
|
|
|
|
|
@property
|
|
|
|
def nO(self):
|
2020-05-18 23:23:33 +03:00
|
|
|
if self.attrs["has_upper"]:
|
|
|
|
return self.vec2scores.get_dim("nO")
|
|
|
|
else:
|
|
|
|
return self.state2vec.get_dim("nO")
|
2018-05-15 23:17:29 +03:00
|
|
|
|
2019-02-24 18:41:41 +03:00
|
|
|
def class_is_unseen(self, class_):
|
|
|
|
return self._class_mask[class_]
|
|
|
|
|
|
|
|
def mark_class_unseen(self, class_):
|
|
|
|
self._class_mask[class_] = 0
|
|
|
|
|
|
|
|
def mark_class_seen(self, class_):
|
|
|
|
self._class_mask[class_] = 1
|
|
|
|
|
2020-06-26 20:34:12 +03:00
|
|
|
def get_token_ids(self, states):
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef StateClass state
|
|
|
|
states = [state for state in states if not state.is_final()]
|
|
|
|
cdef np.ndarray ids = numpy.zeros((len(states), self.state2vec.nF),
|
|
|
|
dtype='i', order='C')
|
|
|
|
ids.fill(-1)
|
|
|
|
c_ids = <int*>ids.data
|
|
|
|
for state in states:
|
|
|
|
state.c.set_context_tokens(c_ids, ids.shape[1])
|
|
|
|
c_ids += ids.shape[1]
|
|
|
|
return ids
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def finish_steps(self, golds):
|
2018-05-15 23:17:29 +03:00
|
|
|
# Add a padding vector to the d_tokvecs gradient, so that missing
|
|
|
|
# values don't affect the real gradient.
|
2020-01-29 19:06:46 +03:00
|
|
|
d_tokvecs = self.ops.alloc((self.tokvecs.shape[0]+1, self.tokvecs.shape[1]))
|
2019-11-19 17:54:34 +03:00
|
|
|
# Tells CUDA to block, so our async copies complete.
|
|
|
|
if self.cuda_stream is not None:
|
|
|
|
self.cuda_stream.synchronize()
|
2018-05-15 23:17:29 +03:00
|
|
|
for ids, d_vector, bp_vector in self.backprops:
|
2020-01-29 19:06:46 +03:00
|
|
|
d_state_features = bp_vector((d_vector, ids))
|
2018-05-15 23:17:29 +03:00
|
|
|
ids = ids.flatten()
|
|
|
|
d_state_features = d_state_features.reshape(
|
|
|
|
(ids.size, d_state_features.shape[2]))
|
|
|
|
self.ops.scatter_add(d_tokvecs, ids,
|
|
|
|
d_state_features)
|
|
|
|
# Padded -- see update()
|
2020-01-29 19:06:46 +03:00
|
|
|
self.bp_tokvecs(d_tokvecs[:-1])
|
2018-05-15 23:17:29 +03:00
|
|
|
return d_tokvecs
|
|
|
|
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def step_forward(model: ParserStepModel, states, is_train):
|
|
|
|
token_ids = model.get_token_ids(states)
|
|
|
|
vector, get_d_tokvecs = model.state2vec(token_ids, is_train)
|
2020-05-18 23:23:33 +03:00
|
|
|
if model.attrs["has_upper"]:
|
2020-01-29 19:06:46 +03:00
|
|
|
scores, get_d_vector = model.vec2scores(vector, is_train)
|
|
|
|
else:
|
|
|
|
scores = NumpyOps().asarray(vector)
|
|
|
|
get_d_vector = lambda d_scores: d_scores
|
|
|
|
# If the class is unseen, make sure its score is minimum
|
|
|
|
scores[:, model._class_mask == 0] = numpy.nanmin(scores)
|
|
|
|
|
|
|
|
def backprop_parser_step(d_scores):
|
|
|
|
# Zero vectors for unseen classes
|
|
|
|
d_scores *= model._class_mask
|
|
|
|
d_vector = get_d_vector(d_scores)
|
|
|
|
if isinstance(model.state2vec.ops, CupyOps) \
|
|
|
|
and not isinstance(token_ids, model.state2vec.ops.xp.ndarray):
|
|
|
|
# Move token_ids and d_vector to GPU, asynchronously
|
|
|
|
model.backprops.append((
|
|
|
|
util.get_async(model.cuda_stream, token_ids),
|
|
|
|
util.get_async(model.cuda_stream, d_vector),
|
|
|
|
get_d_tokvecs
|
|
|
|
))
|
|
|
|
else:
|
|
|
|
model.backprops.append((token_ids, d_vector, get_d_tokvecs))
|
|
|
|
return None
|
|
|
|
return scores, backprop_parser_step
|
|
|
|
|
|
|
|
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef class precompute_hiddens:
|
|
|
|
"""Allow a model to be "primed" by pre-computing input features in bulk.
|
|
|
|
|
|
|
|
This is used for the parser, where we want to take a batch of documents,
|
|
|
|
and compute vectors for each (token, position) pair. These vectors can then
|
|
|
|
be reused, especially for beam-search.
|
|
|
|
|
|
|
|
Let's say we're using 12 features for each state, e.g. word at start of
|
|
|
|
buffer, three words on stack, their children, etc. In the normal arc-eager
|
|
|
|
system, a document of length N is processed in 2*N states. This means we'll
|
|
|
|
create 2*N*12 feature vectors --- but if we pre-compute, we only need
|
|
|
|
N*12 vector computations. The saving for beam-search is much better:
|
|
|
|
if we have a beam of k, we'll normally make 2*N*12*K computations --
|
|
|
|
so we can save the factor k. This also gives a nice CPU/GPU division:
|
|
|
|
we can do all our hard maths up front, packed into large multiplications,
|
|
|
|
and do the hard-to-program parsing on the CPU.
|
|
|
|
"""
|
2020-02-27 20:42:27 +03:00
|
|
|
cdef readonly int nF, nO, nP
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef bint _is_synchronized
|
|
|
|
cdef public object ops
|
|
|
|
cdef np.ndarray _features
|
|
|
|
cdef np.ndarray _cached
|
|
|
|
cdef np.ndarray bias
|
|
|
|
cdef object _cuda_stream
|
|
|
|
cdef object _bp_hiddens
|
2019-11-19 17:54:34 +03:00
|
|
|
cdef object activation
|
2018-05-15 23:17:29 +03:00
|
|
|
|
|
|
|
def __init__(self, batch_size, tokvecs, lower_model, cuda_stream=None,
|
2020-01-29 19:06:46 +03:00
|
|
|
activation="maxout", train=False):
|
|
|
|
gpu_cached, bp_features = lower_model(tokvecs, train)
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef np.ndarray cached
|
|
|
|
if not isinstance(gpu_cached, numpy.ndarray):
|
|
|
|
# Note the passing of cuda_stream here: it lets
|
|
|
|
# cupy make the copy asynchronously.
|
|
|
|
# We then have to block before first use.
|
|
|
|
cached = gpu_cached.get(stream=cuda_stream)
|
|
|
|
else:
|
|
|
|
cached = gpu_cached
|
2020-01-29 19:06:46 +03:00
|
|
|
if not isinstance(lower_model.get_param("b"), numpy.ndarray):
|
2020-03-29 01:09:35 +03:00
|
|
|
self.bias = lower_model.get_param("b").get(stream=cuda_stream)
|
2018-05-15 23:17:29 +03:00
|
|
|
else:
|
2020-01-29 19:06:46 +03:00
|
|
|
self.bias = lower_model.get_param("b")
|
2018-05-15 23:17:29 +03:00
|
|
|
self.nF = cached.shape[1]
|
2020-01-29 19:06:46 +03:00
|
|
|
if lower_model.has_dim("nP"):
|
|
|
|
self.nP = lower_model.get_dim("nP")
|
|
|
|
else:
|
|
|
|
self.nP = 1
|
2018-05-15 23:17:29 +03:00
|
|
|
self.nO = cached.shape[2]
|
|
|
|
self.ops = lower_model.ops
|
2019-11-19 17:54:34 +03:00
|
|
|
assert activation in (None, "relu", "maxout")
|
|
|
|
self.activation = activation
|
2018-05-15 23:17:29 +03:00
|
|
|
self._is_synchronized = False
|
|
|
|
self._cuda_stream = cuda_stream
|
|
|
|
self._cached = cached
|
|
|
|
self._bp_hiddens = bp_features
|
|
|
|
|
|
|
|
cdef const float* get_feat_weights(self) except NULL:
|
|
|
|
if not self._is_synchronized and self._cuda_stream is not None:
|
|
|
|
self._cuda_stream.synchronize()
|
|
|
|
self._is_synchronized = True
|
|
|
|
return <float*>self._cached.data
|
|
|
|
|
2020-02-27 20:42:27 +03:00
|
|
|
def has_dim(self, name):
|
|
|
|
if name == "nF":
|
|
|
|
return self.nF if self.nF is not None else True
|
|
|
|
elif name == "nP":
|
|
|
|
return self.nP if self.nP is not None else True
|
|
|
|
elif name == "nO":
|
|
|
|
return self.nO if self.nO is not None else True
|
|
|
|
else:
|
|
|
|
return False
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def get_dim(self, name):
|
|
|
|
if name == "nF":
|
|
|
|
return self.nF
|
|
|
|
elif name == "nP":
|
|
|
|
return self.nP
|
|
|
|
elif name == "nO":
|
|
|
|
return self.nO
|
|
|
|
else:
|
|
|
|
raise ValueError(f"Dimension {name} invalid -- only nO, nF, nP")
|
|
|
|
|
2020-02-27 20:42:27 +03:00
|
|
|
def set_dim(self, name, value):
|
|
|
|
if name == "nF":
|
|
|
|
self.nF = value
|
|
|
|
elif name == "nP":
|
|
|
|
self.nP = value
|
|
|
|
elif name == "nO":
|
|
|
|
self.nO = value
|
|
|
|
else:
|
|
|
|
raise ValueError(f"Dimension {name} invalid -- only nO, nF, nP")
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def __call__(self, X, bint is_train):
|
|
|
|
if is_train:
|
|
|
|
return self.begin_update(X)
|
|
|
|
else:
|
|
|
|
return self.predict(X), lambda X: X
|
|
|
|
|
|
|
|
def predict(self, X):
|
|
|
|
return self.begin_update(X)[0]
|
2018-05-15 23:17:29 +03:00
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def begin_update(self, token_ids):
|
2018-05-15 23:17:29 +03:00
|
|
|
cdef np.ndarray state_vector = numpy.zeros(
|
|
|
|
(token_ids.shape[0], self.nO, self.nP), dtype='f')
|
|
|
|
# This is tricky, but (assuming GPU available);
|
|
|
|
# - Input to forward on CPU
|
|
|
|
# - Output from forward on CPU
|
|
|
|
# - Input to backward on GPU!
|
|
|
|
# - Output from backward on GPU
|
|
|
|
bp_hiddens = self._bp_hiddens
|
|
|
|
|
|
|
|
feat_weights = self.get_feat_weights()
|
|
|
|
cdef int[:, ::1] ids = token_ids
|
|
|
|
sum_state_features(<float*>state_vector.data,
|
|
|
|
feat_weights, &ids[0,0],
|
|
|
|
token_ids.shape[0], self.nF, self.nO*self.nP)
|
2020-01-29 19:06:46 +03:00
|
|
|
state_vector = state_vector + self.bias
|
2018-05-15 23:17:29 +03:00
|
|
|
state_vector, bp_nonlinearity = self._nonlinearity(state_vector)
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def backward(d_state_vector_ids):
|
2018-05-15 23:17:29 +03:00
|
|
|
d_state_vector, token_ids = d_state_vector_ids
|
2020-01-29 19:06:46 +03:00
|
|
|
d_state_vector = bp_nonlinearity(d_state_vector)
|
|
|
|
d_tokens = bp_hiddens((d_state_vector, token_ids))
|
2018-05-15 23:17:29 +03:00
|
|
|
return d_tokens
|
|
|
|
return state_vector, backward
|
|
|
|
|
|
|
|
def _nonlinearity(self, state_vector):
|
2019-10-28 01:43:09 +03:00
|
|
|
if isinstance(state_vector, numpy.ndarray):
|
|
|
|
ops = NumpyOps()
|
|
|
|
else:
|
|
|
|
ops = CupyOps()
|
2019-12-22 03:53:56 +03:00
|
|
|
|
2019-11-19 17:54:34 +03:00
|
|
|
if self.activation == "maxout":
|
2019-10-28 01:43:09 +03:00
|
|
|
state_vector, mask = ops.maxout(state_vector)
|
2019-11-19 17:54:34 +03:00
|
|
|
else:
|
|
|
|
state_vector = state_vector.reshape(state_vector.shape[:-1])
|
|
|
|
if self.activation == "relu":
|
|
|
|
mask = state_vector >= 0.
|
|
|
|
state_vector *= mask
|
|
|
|
else:
|
|
|
|
mask = None
|
2018-05-15 23:17:29 +03:00
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
def backprop_nonlinearity(d_best):
|
2019-10-28 01:43:09 +03:00
|
|
|
if isinstance(d_best, numpy.ndarray):
|
|
|
|
ops = NumpyOps()
|
|
|
|
else:
|
|
|
|
ops = CupyOps()
|
2019-11-19 17:54:34 +03:00
|
|
|
if mask is not None:
|
|
|
|
mask_ = ops.asarray(mask)
|
2019-10-28 01:43:09 +03:00
|
|
|
# This will usually be on GPU
|
|
|
|
d_best = ops.asarray(d_best)
|
2018-12-20 18:12:22 +03:00
|
|
|
# Fix nans (which can occur from unseen classes.)
|
2020-05-20 13:27:31 +03:00
|
|
|
try:
|
|
|
|
d_best[ops.xp.isnan(d_best)] = 0.
|
|
|
|
except:
|
|
|
|
print(ops.xp.isnan(d_best))
|
|
|
|
raise
|
2019-11-19 17:54:34 +03:00
|
|
|
if self.activation == "maxout":
|
|
|
|
mask_ = ops.asarray(mask)
|
|
|
|
return ops.backprop_maxout(d_best, mask_, self.nP)
|
|
|
|
elif self.activation == "relu":
|
|
|
|
mask_ = ops.asarray(mask)
|
2019-10-28 01:43:09 +03:00
|
|
|
d_best *= mask_
|
2018-05-15 23:17:29 +03:00
|
|
|
d_best = d_best.reshape((d_best.shape + (1,)))
|
|
|
|
return d_best
|
|
|
|
else:
|
2019-11-19 17:54:34 +03:00
|
|
|
return d_best.reshape((d_best.shape + (1,)))
|
2018-05-15 23:17:29 +03:00
|
|
|
return state_vector, backprop_nonlinearity
|