# -*- encoding: utf-8 -*-
# pylint: disable=E0203,E1101,C0111
"""
Runtime operator.
:githublink:`%|py|7`
"""
import numpy
from ..shape_object import ShapeObject
from ._op import OpRun
[docs]class LabelEncoder(OpRun):
atts = {'default_float': 0., 'default_int64': -1,
'default_string': b'',
'keys_floats': numpy.empty(0, dtype=numpy.float32),
'keys_int64s': numpy.empty(0, dtype=numpy.int64),
'keys_strings': numpy.empty(0, dtype=numpy.str),
'values_floats': numpy.empty(0, dtype=numpy.float32),
'values_int64s': numpy.empty(0, dtype=numpy.int64),
'values_strings': numpy.empty(0, dtype=numpy.str),
}
[docs] def __init__(self, onnx_node, desc=None, **options):
OpRun.__init__(self, onnx_node, desc=desc,
expected_attributes=LabelEncoder.atts,
**options)
if len(self.keys_floats) > 0 and len(self.values_floats) > 0:
self.classes_ = {k: v for k, v in zip(
self.keys_floats, self.values_floats)}
self.default_ = self.default_float
self.dtype_ = numpy.float32
elif len(self.keys_floats) > 0 and len(self.values_int64s) > 0:
self.classes_ = {k: v for k, v in zip(
self.keys_floats, self.values_int64s)}
self.default_ = self.default_int64
self.dtype_ = numpy.int64
elif len(self.keys_int64s) > 0 and len(self.values_int64s) > 0:
self.classes_ = {k: v for k, v in zip(
self.keys_int64s, self.values_int64s)}
self.default_ = self.default_int64
self.dtype_ = numpy.int64
elif len(self.keys_int64s) > 0 and len(self.values_floats) > 0:
self.classes_ = {k: v for k, v in zip(
self.keys_int64s, self.values_floats)}
self.default_ = self.default_int64
self.dtype_ = numpy.float32
elif len(self.keys_strings) > 0 and len(self.values_int64s) > 0:
self.classes_ = {k.decode('utf-8'): v for k, v in zip(
self.keys_strings, self.values_int64s)}
self.default_ = self.default_int64
self.dtype_ = numpy.int64
elif len(self.keys_strings) > 0 and len(self.values_strings) > 0:
self.classes_ = {
k.decode('utf-8'): v.decode('utf-8') for k, v in zip(
self.keys_strings, self.values_strings)}
self.default_ = self.default_string
self.dtype_ = numpy.array(self.classes_.values).dtype
elif len(self.keys_floats) > 0 and len(self.values_strings) > 0:
self.classes_ = {k: v.decode('utf-8') for k, v in zip(
self.keys_floats, self.values_strings)}
self.default_ = self.default_string
self.dtype_ = numpy.array(self.classes_.values).dtype
elif hasattr(self, 'classes_strings'):
raise RuntimeError( # pragma: no cover
"This runtime does not implement version 1 of "
"operator LabelEncoder.")
else:
raise RuntimeError(
"No encoding was defined in {}.".format(onnx_node))
if len(self.classes_) == 0:
raise RuntimeError( # pragma: no cover
"Empty classes for LabelEncoder, (onnx_node='{}')\n{}.".format(
self.onnx_node.name, onnx_node))
[docs] def _run(self, x): # pylint: disable=W0221
if len(x.shape) > 1:
x = numpy.squeeze(x)
res = numpy.empty((x.shape[0], ), dtype=self.dtype_)
for i in range(0, res.shape[0]):
res[i] = self.classes_.get(x[i], self.default_)
return (res, )
[docs] def _infer_shapes(self, x): # pylint: disable=W0221
nb = len(self.classes_.values())
return (ShapeObject((x[0], nb), dtype=self.dtype_,
name="{}-1".format(self.__class__.__name__)), )
