explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-02-09 16:10:33 +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:
Matthew Honnibal 2018-12-10 09:44:07 +01:00
commit 6936ca1664
2 changed files with 7 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
spacy/cli/train.py
View File

@ -9,6 +9,7 @@ from timeit import default_timer as timer
import shutil import shutil
import srsly import srsly
from wasabi import Printer from wasabi import Printer
from thinc.rates import slanted_triangular
from .._ml import create_default_optimizer from .._ml import create_default_optimizer
from ..attrs import PROB, IS_OOV, CLUSTER, LANG from ..attrs import PROB, IS_OOV, CLUSTER, LANG
@ -118,7 +119,7 @@ def train(
) )
batch_sizes = util.compounding( batch_sizes = util.compounding(
util.env_opt("batch_from", 100.0), util.env_opt("batch_from", 100.0),
util.env_opt("batch_to", 1000.0), util.env_opt("batch_to", 2000.0),
util.env_opt("batch_compound", 1.001), util.env_opt("batch_compound", 1.001),
) )
@ -180,6 +181,8 @@ def train(
# Start with a blank model, call begin_training # Start with a blank model, call begin_training
optimizer = nlp.begin_training(lambda: corpus.train_tuples, device=use_gpu) optimizer = nlp.begin_training(lambda: corpus.train_tuples, device=use_gpu)
optimizer.b1_decay = 0.0001
optimizer.b2_decay = 0.0001
nlp._optimizer = None nlp._optimizer = None
# Load in pre-trained weights # Load in pre-trained weights

5
spacy/syntax/ner.pyx
View File

@ -207,15 +207,16 @@ cdef class BiluoPushDown(TransitionSystem):
else: else:
label_id = label_name label_id = label_name
if action == OUT and label_id != 0: if action == OUT and label_id != 0:
return return None
if action == MISSING or action == ISNT: if action == MISSING or action == ISNT:
return return None
# Check we're not creating a move we already have, so that this is # Check we're not creating a move we already have, so that this is
# idempotent # idempotent
for trans in self.c[:self.n_moves]: for trans in self.c[:self.n_moves]:
if trans.move == action and trans.label == label_id: if trans.move == action and trans.label == label_id:
return 0 return 0
if self.n_moves >= self._size: if self.n_moves >= self._size:
self._size = self.n_moves
self._size *= 2 self._size *= 2
self.c = <Transition*>self.mem.realloc(self.c, self._size * sizeof(self.c[0])) self.c = <Transition*>self.mem.realloc(self.c, self._size * sizeof(self.c[0]))
self.c[self.n_moves] = self.init_transition(self.n_moves, action, label_id) self.c[self.n_moves] = self.init_transition(self.n_moves, action, label_id)