def depthwise_conv3d_args_preprocessor(args, kwargs):
converted = []
if 'init' in kwargs:
init = kwargs.pop('init')
kwargs['depthwise_initializer'] = init
converted.append(('init', 'depthwise_initializer'))
args, kwargs, _converted = conv3d_args_preprocessor(args, kwargs)
return args, kwargs, converted + _converted
legacy_depthwise_conv3d_support = generate_legacy_interface(
allowed_positional_args=['filters', 'kernel_size'],
conversions=[('nb_filter', 'filters'), ('subsample', 'strides'),
('border_mode', 'padding'),
('dim_ordering', 'data_format'),
('b_regularizer', 'bias_regularizer'),
('b_constraint', 'bias_constraint'),
('bias', 'use_bias')],
value_conversions={
'dim_ordering': {
'tf': 'channels_last',
'th': 'channels_first',
'default': None
}
},
preprocessor=depthwise_conv3d_args_preprocessor)
def depthwise_conv2d_args_preprocessor(args, kwargs):
converted = []
if 'init' in kwargs:
init = kwargs.pop('init')
kwargs['depthwise_initializer'] = init
converted.append(('init', 'depthwise_initializer'))
args, kwargs, _converted = conv2d_args_preprocessor(args, kwargs)
return args, kwargs, converted + _converted
legacy_depthwise_conv2d_support = generate_legacy_interface(
allowed_positional_args=['filters', 'kernel_size'],
conversions=[('nb_filter', 'filters'),
('subsample', 'strides'),
('border_mode', 'padding'),
('dim_ordering', 'data_format'),
('b_regularizer', 'bias_regularizer'),
('b_constraint', 'bias_constraint'),
('bias', 'use_bias')],
value_conversions={'dim_ordering': {'tf': 'channels_last',
'th': 'channels_first',
'default': None}},
preprocessor=depthwise_conv2d_args_preprocessor)
class DepthwiseConv2D(Conv2D):
@legacy_depthwise_conv2d_support
def __init__(self, filters,
kernel_size,
strides=(1, 1),
padding='valid',
import os
import numpy as np
from keras import backend as K
from keras.legacy.interfaces import generate_legacy_interface, recurrent_args_preprocessor
from keras.models import model_from_json
legacy_prednet_support = generate_legacy_interface(
allowed_positional_args=[
'stack_sizes', 'R_stack_sizes', 'A_filt_sizes', 'Ahat_filt_sizes',
'R_filt_sizes'
],
conversions=[('dim_ordering', 'data_format'),
('consume_less', 'implementation')],
value_conversions={
'dim_ordering': {
'tf': 'channels_last',
'th': 'channels_first',
'default': None
},
'consume_less': {
'cpu': 0,
'mem': 1,
'gpu': 2
}
},
preprocessor=recurrent_args_preprocessor)
# Convert old Keras (1.2) json models and weights to Keras 2.0
def convert_model_to_keras2(old_json_file, old_weights_file, new_json_file,
new_weights_file):
from prednet import PredNet
from keras.engine import InputSpec
from keras import backend as K
from keras.utils.generic_utils import to_list
from keras.utils.generic_utils import deserialize_keras_object
from keras.legacy.interfaces import generate_legacy_interface
from scipy.sparse import csr_matrix, issparse
from functools import partial
from .backend_keras import is_tensor, dumps_sparse, loads_sparse
from .backend_keras import make_sparse, sp_floor
from .utils import patch_assert_input_compatibility
from .utils import patch_predict_model
import numpy as np
legacy_entitie_support = generate_legacy_interface(
allowed_positional_args=['units'], conversions=[('output_dim', 'units')])
# esto solo aporta claridad a los diagramas de las redes ya que todo queda encapsulado
# como una clase de keras y se ve como un nodo en el grafo, no es m\'as que una capa
# densa por cada entidad(2 entidades) y otra capa densa mas grande conectada al concat
# de las capas antes mencionadas
@patch_assert_input_compatibility
class RelationLayer(Layer):
@legacy_entitie_support
def __init__(self, units, **kwargs):
if 'input_shape' not in kwargs and 'input_dim' in kwargs:
kwargs['input_shape'] = (kwargs.pop('input_dim'), )
super(RelationLayer, self).__init__(**kwargs)
self.input_spec = InputSpec(min_ndim=2)
self.units = units
