spaCy/spacy/syntax/transition_system.pyx

200 lines
6.6 KiB
Cython
Raw Normal View History

2016-10-27 19:08:13 +03:00
# cython: infer_types=True
2017-04-15 14:05:15 +03:00
# coding: utf-8
from __future__ import unicode_literals
2017-10-27 20:45:57 +03:00
from cpython.ref cimport Py_INCREF
from cymem.cymem cimport Pool
from thinc.typedefs cimport weight_t
from collections import OrderedDict, Counter
import ujson
from ..structs cimport TokenC
from .stateclass cimport StateClass
2017-05-28 19:09:27 +03:00
from ..typedefs cimport attr_t
2017-10-27 20:45:57 +03:00
from ..compat import json_dumps
from ..errors import Errors
2017-10-27 20:45:57 +03:00
from .. import util
cdef weight_t MIN_SCORE = -90000
class OracleError(Exception):
pass
2017-03-11 20:12:01 +03:00
cdef void* _init_state(Pool mem, int length, void* tokens) except NULL:
2017-11-14 04:11:40 +03:00
cdef StateC* st = new StateC(<const TokenC*>tokens, length)
2017-03-11 20:12:01 +03:00
return <void*>st
cdef class TransitionSystem:
def __init__(self, StringStore string_table, labels_by_action=None, min_freq=None):
self.mem = Pool()
self.strings = string_table
self.n_moves = 0
self._size = 100
2017-03-11 20:12:01 +03:00
self.c = <Transition*>self.mem.alloc(self._size, sizeof(Transition))
2017-03-11 20:12:01 +03:00
self.labels = {}
if labels_by_action:
self.initialize_actions(labels_by_action, min_freq=min_freq)
2017-05-28 19:09:27 +03:00
self.root_label = self.strings.add('ROOT')
2017-03-11 20:12:01 +03:00
self.init_beam_state = _init_state
def __reduce__(self):
return (self.__class__, (self.strings, self.labels), None, None)
def init_batch(self, docs):
cdef StateClass state
states = []
offset = 0
for doc in docs:
state = StateClass(doc, offset=offset)
self.initialize_state(state.c)
states.append(state)
offset += len(doc)
return states
2017-05-26 19:31:23 +03:00
def get_oracle_sequence(self, doc, GoldParse gold):
cdef Pool mem = Pool()
costs = <float*>mem.alloc(self.n_moves, sizeof(float))
is_valid = <int*>mem.alloc(self.n_moves, sizeof(int))
cdef StateClass state = StateClass(doc, offset=0)
self.initialize_state(state.c)
history = []
while not state.is_final():
self.set_costs(is_valid, costs, state, gold)
for i in range(self.n_moves):
if is_valid[i] and costs[i] <= 0:
action = self.c[i]
history.append(i)
action.do(state.c, action.label)
break
else:
raise ValueError(Errors.E024)
2017-05-26 19:31:23 +03:00
return history
cdef int initialize_state(self, StateC* state) nogil:
pass
cdef int finalize_state(self, StateC* state) nogil:
pass
def finalize_doc(self, doc):
pass
def preprocess_gold(self, GoldParse gold):
raise NotImplementedError
def is_gold_parse(self, StateClass state, GoldParse gold):
raise NotImplementedError
cdef Transition lookup_transition(self, object name) except *:
raise NotImplementedError
2017-05-28 15:06:40 +03:00
cdef Transition init_transition(self, int clas, int move, attr_t label) except *:
raise NotImplementedError
def is_valid(self, StateClass stcls, move_name):
action = self.lookup_transition(move_name)
if action.move == 0:
return False
return action.is_valid(stcls.c, action.label)
cdef int set_valid(self, int* is_valid, const StateC* st) nogil:
2015-06-04 20:32:32 +03:00
cdef int i
for i in range(self.n_moves):
is_valid[i] = self.c[i].is_valid(st, self.c[i].label)
2015-06-04 20:32:32 +03:00
cdef int set_costs(self, int* is_valid, weight_t* costs,
StateClass stcls, GoldParse gold) except -1:
cdef int i
self.set_valid(is_valid, stcls.c)
2017-05-26 19:31:23 +03:00
cdef int n_gold = 0
for i in range(self.n_moves):
if is_valid[i]:
costs[i] = self.c[i].get_cost(stcls, &gold.c, self.c[i].label)
2017-05-26 19:31:23 +03:00
n_gold += costs[i] <= 0
else:
costs[i] = 9000
2017-05-26 19:31:23 +03:00
if n_gold <= 0:
raise ValueError(Errors.E024)
2017-05-28 19:09:27 +03:00
def get_class_name(self, int clas):
act = self.c[clas]
return self.move_name(act.move, act.label)
def initialize_actions(self, labels_by_action, min_freq=None):
self.labels = {}
self.n_moves = 0
for action, label_freqs in sorted(labels_by_action.items()):
action = int(action)
# Make sure we take a copy here, and that we get a Counter
self.labels[action] = Counter()
# Have to be careful here: Sorting must be stable, or our model
# won't be read back in correctly.
sorted_labels = [(f, L) for L, f in label_freqs.items()]
sorted_labels.sort()
sorted_labels.reverse()
for freq, label_str in sorted_labels:
self.add_action(int(action), label_str)
self.labels[action][label_str] = freq
2017-05-28 19:09:27 +03:00
def add_action(self, int action, label_name):
cdef attr_t label_id
2017-10-18 22:48:00 +03:00
if not isinstance(label_name, int) and \
2017-10-27 20:45:57 +03:00
not isinstance(label_name, long):
2017-05-28 19:09:27 +03:00
label_id = self.strings.add(label_name)
else:
label_id = label_name
# Check we're not creating a move we already have, so that this is
# idempotent
for trans in self.c[:self.n_moves]:
2017-05-28 19:09:27 +03:00
if trans.move == action and trans.label == label_id:
return 0
if self.n_moves >= self._size:
self._size *= 2
self.c = <Transition*>self.mem.realloc(self.c, self._size * sizeof(self.c[0]))
2017-05-28 19:09:27 +03:00
self.c[self.n_moves] = self.init_transition(self.n_moves, action, label_id)
self.n_moves += 1
if self.labels.get(action, []):
new_freq = min(self.labels[action].values())
else:
self.labels[action] = Counter()
new_freq = -1
if new_freq > 0:
new_freq = 0
self.labels[action][label_name] = new_freq-1
return 1
def to_disk(self, path, **exclude):
2017-05-31 14:44:00 +03:00
with path.open('wb') as file_:
file_.write(self.to_bytes(**exclude))
def from_disk(self, path, **exclude):
2017-05-31 14:44:00 +03:00
with path.open('rb') as file_:
byte_data = file_.read()
self.from_bytes(byte_data, **exclude)
return self
def to_bytes(self, **exclude):
transitions = []
serializers = {
'moves': lambda: json_dumps(self.labels),
'strings': lambda: self.strings.to_bytes()
}
return util.to_bytes(serializers, exclude)
def from_bytes(self, bytes_data, **exclude):
labels = {}
deserializers = {
'moves': lambda b: labels.update(ujson.loads(b)),
'strings': lambda b: self.strings.from_bytes(b)
}
msg = util.from_bytes(bytes_data, deserializers, exclude)
self.initialize_actions(labels)
return self