explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-07-15 02:32:37 +03:00
Code Issues Packages Projects Releases Wiki Activity

bool -> float

Browse Source
This commit is contained in:
vinit 2023-02-20 16:54:37 +05:30
parent b0929271a8
commit ee5332eba5
3 changed files with 11 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
spacy/cli/debug_data.py
View File

@ -7,7 +7,7 @@ import srsly
from wasabi import Printer, MESSAGES, msg from wasabi import Printer, MESSAGES, msg
import typer import typer
import math import math
import numpy as np import numpy
from ._util import app, Arg, Opt, show_validation_error, parse_config_overrides from ._util import app, Arg, Opt, show_validation_error, parse_config_overrides
from ._util import import_code, debug_cli, _format_number from ._util import import_code, debug_cli, _format_number
@ -524,9 +524,9 @@ def debug_data(
msg.divider("Part-of-speech Tagging") msg.divider("Part-of-speech Tagging")
label_list, counts = zip(*gold_train_data["tags"].items()) label_list, counts = zip(*gold_train_data["tags"].items())
msg.info(f"{len(label_list)} label(s) in train data") msg.info(f"{len(label_list)} label(s) in train data")
p = np.array(counts) p = numpy.array(counts)
p = p / p.sum() p = p / p.sum()
norm_entropy = (-p * np.log2(p)).sum() / np.log2(len(label_list)) norm_entropy = (-p * numpy.log2(p)).sum() / numpy.log2(len(label_list))
msg.info(f"{norm_entropy} is the normalised label entropy") msg.info(f"{norm_entropy} is the normalised label entropy")
model_labels = _get_labels_from_model(nlp, "tagger") model_labels = _get_labels_from_model(nlp, "tagger")
labels = set(label_list) labels = set(label_list)

7
spacy/pipeline/tagger.pyx
View File

@ -45,7 +45,7 @@ DEFAULT_TAGGER_MODEL = Config().from_str(default_model_config)["model"]
@Language.factory( @Language.factory(
"tagger", "tagger",
assigns=["token.tag"], assigns=["token.tag"],
default_config={"model": DEFAULT_TAGGER_MODEL, "overwrite": False, "scorer": {"@scorers": "spacy.tagger_scorer.v1"}, "neg_prefix": "!", "label_smoothing": False}, default_config={"model": DEFAULT_TAGGER_MODEL, "overwrite": False, "scorer": {"@scorers": "spacy.tagger_scorer.v1"}, "neg_prefix": "!", "label_smoothing": 0.05},
default_score_weights={"tag_acc": 1.0}, default_score_weights={"tag_acc": 1.0},
) )
def make_tagger( def make_tagger(
@ -55,7 +55,7 @@ def make_tagger(
overwrite: bool, overwrite: bool,
scorer: Optional[Callable], scorer: Optional[Callable],
neg_prefix: str, neg_prefix: str,
label_smoothing: bool, label_smoothing: float,
): ):
"""Construct a part-of-speech tagger component. """Construct a part-of-speech tagger component.
@ -90,7 +90,7 @@ class Tagger(TrainablePipe):
overwrite=BACKWARD_OVERWRITE, overwrite=BACKWARD_OVERWRITE,
scorer=tagger_score, scorer=tagger_score,
neg_prefix="!", neg_prefix="!",
label_smoothing=False, label_smoothing=0.05,
): ):
"""Initialize a part-of-speech tagger. """Initialize a part-of-speech tagger.
@ -258,7 +258,6 @@ class Tagger(TrainablePipe):
DOCS: https://spacy.io/api/tagger#get_loss DOCS: https://spacy.io/api/tagger#get_loss
""" """
validate_examples(examples, "Tagger.get_loss") validate_examples(examples, "Tagger.get_loss")
self.cfg["label_smoothing"] = 0.05 if self.cfg["label_smoothing"] else 0.0
loss_func = SequenceCategoricalCrossentropy(names=self.labels, normalize=False, neg_prefix=self.cfg["neg_prefix"], label_smoothing=self.cfg["label_smoothing"]) loss_func = SequenceCategoricalCrossentropy(names=self.labels, normalize=False, neg_prefix=self.cfg["neg_prefix"], label_smoothing=self.cfg["label_smoothing"])
# Convert empty tag "" to missing value None so that both misaligned # Convert empty tag "" to missing value None so that both misaligned
# tokens and tokens with missing annotation have the default missing # tokens and tokens with missing annotation have the default missing

10
spacy/tests/pipeline/test_tagger.py
View File

@ -70,10 +70,10 @@ PARTIAL_DATA = [
def test_label_smoothing(): def test_label_smoothing():
nlp = Language() nlp = Language()
tagger_no_ls = nlp.add_pipe( tagger_no_ls = nlp.add_pipe(
"tagger", "no_label_smoothing", config=dict(label_smoothing=False) "tagger", "no_label_smoothing", config=dict(label_smoothing=0.0)
) )
tagger_ls = nlp.add_pipe( tagger_ls = nlp.add_pipe(
"tagger", "label_smoothing", config=dict(label_smoothing=True) "tagger", "label_smoothing"
) )
train_examples = [] train_examples = []
losses = {} losses = {}
@ -87,9 +87,9 @@ def test_label_smoothing():
tag_scores, bp_tag_scores = tagger_ls.model.begin_update( tag_scores, bp_tag_scores = tagger_ls.model.begin_update(
[eg.predicted for eg in train_examples] [eg.predicted for eg in train_examples]
) )
no_ls_probs = tagger_no_ls.get_loss(train_examples, tag_scores)[1][0] no_ls_grads= tagger_no_ls.get_loss(train_examples, tag_scores)[1][0]
ls_probs = tagger_ls.get_loss(train_examples, tag_scores)[1][0] ls_grads= tagger_ls.get_loss(train_examples, tag_scores)[1][0]
assert_array_almost_equal((ls_probs - no_ls_probs)[0], [0.05, -0.025, -0.025]) assert_array_almost_equal((ls_grads - no_ls_grads)[0], [0.05, -0.025, -0.025])
def test_no_label(): def test_no_label():