import numpy as np
from malaya_speech.model.frame import Frame
from malaya_speech.utils.padding import (
sequence_nd as padding_sequence_nd,
sequence_1d,
)
from malaya_speech.utils.activation import softmax
from malaya_speech.model.abstract import Abstract
class Transformer2Vec(Abstract):
def __init__(
self,
input_nodes,
output_nodes,
sess,
model,
label,
name,
vectorizer=None,
extra=None,
):
self._input_nodes = input_nodes
self._output_nodes = output_nodes
self._sess = sess
self.labels = label
self.__model__ = model
self.__name__ = name
def vectorize(self, inputs):
"""
Vectorize inputs.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: np.array
returned [B, D].
"""
inputs = [
input.array if isinstance(input, Frame) else input
for input in inputs
]
inputs, lengths = padding_sequence_nd(
inputs, dim=0, return_len=True
)
r = self._execute(
inputs=[inputs, lengths],
input_labels=['X_placeholder', 'X_len_placeholder'],
output_labels=['logits'],
)
return r['logits']
def __call__(self, inputs):
return self.vectorize(inputs)
[docs]class Speakernet(Abstract):
def __init__(
self,
input_nodes,
output_nodes,
vectorizer,
sess,
model,
extra,
label,
name,
):
self._input_nodes = input_nodes
self._output_nodes = output_nodes
self._vectorizer = vectorizer
self._sess = sess
self._extra = extra
self.labels = label
self.__model__ = model
self.__name__ = name
[docs] def vectorize(self, inputs):
"""
Vectorize inputs.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: np.array
returned [B, D].
"""
inputs = [
input.array if isinstance(input, Frame) else input
for input in inputs
]
inputs = [self._vectorizer(input) for input in inputs]
inputs, lengths = padding_sequence_nd(
inputs, dim=0, return_len=True
)
r = self._execute(
inputs=[inputs, lengths],
input_labels=['Placeholder', 'Placeholder_1'],
output_labels=['logits'],
)
return r['logits']
def __call__(self, inputs):
return self.vectorize(inputs)
[docs]class Speaker2Vec(Abstract):
def __init__(
self,
input_nodes,
output_nodes,
vectorizer,
sess,
model,
extra,
label,
name,
):
self._input_nodes = input_nodes
self._output_nodes = output_nodes
self._vectorizer = vectorizer
self._sess = sess
self._extra = extra
self.labels = label
self.__model__ = model
self.__name__ = name
[docs] def vectorize(self, inputs):
"""
Vectorize inputs.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: np.array
returned [B, D].
"""
inputs = [
input.array if isinstance(input, Frame) else input
for input in inputs
]
inputs = [self._vectorizer(input, **self._extra) for input in inputs]
if self.__model__ == 'deep-speaker':
dim = 0
else:
dim = 1
inputs = padding_sequence_nd(inputs, dim=dim)
inputs = np.expand_dims(inputs, -1)
r = self._execute(
inputs=[inputs],
input_labels=['Placeholder'],
output_labels=['logits'],
)
return r['logits']
def __call__(self, inputs):
return self.vectorize(inputs)
[docs]class SpeakernetClassification(Abstract):
def __init__(
self,
input_nodes,
output_nodes,
vectorizer,
sess,
model,
extra,
label,
name,
):
self._input_nodes = input_nodes
self._output_nodes = output_nodes
self._vectorizer = vectorizer
self._sess = sess
self._extra = extra
self.labels = label
self.__model__ = model
self.__name__ = name
[docs] def predict_proba(self, inputs):
"""
Predict inputs, will return probability.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: np.array
returned [B, D].
"""
inputs = [
input.array if isinstance(input, Frame) else input
for input in inputs
]
inputs = [self._vectorizer(input) for input in inputs]
inputs, lengths = padding_sequence_nd(
inputs, dim=0, return_len=True
)
r = self._execute(
inputs=[inputs, lengths],
input_labels=['Placeholder', 'Placeholder_1'],
output_labels=['logits'],
)
return softmax(r['logits'], axis=-1)
[docs] def predict(self, inputs):
"""
Predict inputs, will return labels.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: List[str]
returned [B].
"""
probs = np.argmax(self.predict_proba(inputs), axis=1)
return [self.labels[p] for p in probs]
def __call__(self, input):
"""
Predict input, will return label.
Parameters
----------
inputs: np.array
np.array or malaya_speech.model.frame.Frame.
Returns
-------
result: str
"""
return self.predict([input])[0]
[docs]class Classification(Abstract):
def __init__(
self,
input_nodes,
output_nodes,
vectorizer,
sess,
model,
extra,
label,
name,
):
self._input_nodes = input_nodes
self._output_nodes = output_nodes
self._vectorizer = vectorizer
self._sess = sess
self._extra = extra
self.labels = label
self.__model__ = model
self.__name__ = name
[docs] def predict_proba(self, inputs):
"""
Predict inputs, will return probability.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: np.array
returned [B, D].
"""
inputs = [
input.array if isinstance(input, Frame) else input
for input in inputs
]
inputs = [self._vectorizer(input, **self._extra) for input in inputs]
if self.__model__ == 'deep-speaker':
dim = 0
else:
dim = 1
inputs = padding_sequence_nd(inputs, dim=dim)
inputs = np.expand_dims(inputs, -1)
r = self._execute(
inputs=[inputs],
input_labels=['Placeholder'],
output_labels=['logits'],
)
return softmax(r['logits'], axis=-1)
[docs] def predict(self, inputs):
"""
Predict inputs, will return labels.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: List[str]
returned [B].
"""
probs = np.argmax(self.predict_proba(inputs), axis=1)
return [self.labels[p] for p in probs]
def __call__(self, input):
"""
Predict input, will return label.
Parameters
----------
inputs: np.array
np.array or malaya_speech.model.frame.Frame.
Returns
-------
result: str
"""
return self.predict([input])[0]
class MarbleNetClassification(Abstract):
def __init__(
self,
input_nodes,
output_nodes,
vectorizer,
sess,
model,
extra,
label,
name,
):
self._input_nodes = input_nodes
self._output_nodes = output_nodes
self._vectorizer = vectorizer
self._sess = sess
self._extra = extra
self.labels = label
self.__model__ = model
self.__name__ = name
def predict_proba(self, inputs):
"""
Predict inputs, will return probability.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: np.array
returned [B, D].
"""
inputs = [
input.array if isinstance(input, Frame) else input
for input in inputs
]
inputs, lengths = sequence_1d(
inputs, return_len=True
)
r = self._execute(
inputs=[inputs, lengths],
input_labels=['X_placeholder', 'X_len_placeholder'],
output_labels=['logits'],
)
return softmax(r['logits'], axis=-1)
def predict(self, inputs):
"""
Predict inputs, will return labels.
Parameters
----------
inputs: List[np.array]
List[np.array] or List[malaya_speech.model.frame.Frame].
Returns
-------
result: List[str]
returned [B].
"""
probs = np.argmax(self.predict_proba(inputs), axis=1)
return [self.labels[p] for p in probs]
def __call__(self, input):
"""
Predict input, will return label.
Parameters
----------
inputs: np.array
np.array or malaya_speech.model.frame.Frame.
Returns
-------
result: str
"""
return self.predict([input])[0]
