def __init__(self,
input_size,
batch_size=64,
hidden_size=512,
num_classes=11):
super(StackedRNNForCPU, self).__init__()
self.batch_size = batch_size
self.input_size = input_size
k = (1 / hidden_size)**0.5
self.w1 = Parameter(
np.random.uniform(-k, k,
(4 * hidden_size *
(input_size + hidden_size + 1), 1, 1)).astype(
np.float32))
self.w2 = Parameter(
np.random.uniform(-k, k, (4 * hidden_size * (2 * hidden_size + 1),
1, 1)).astype(np.float32))
self.h = Tensor(
np.zeros(shape=(1, batch_size, hidden_size)).astype(np.float32))
self.c = Tensor(
np.zeros(shape=(1, batch_size, hidden_size)).astype(np.float32))
self.lstm_1 = nn.LSTMCell(input_size=input_size,
hidden_size=hidden_size)
self.lstm_2 = nn.LSTMCell(input_size=hidden_size,
hidden_size=hidden_size)
self.fc = nn.Dense(in_channels=hidden_size, out_channels=num_classes)
self.transpose = P.Transpose()
def __init__(self,
input_size,
hidden_size,
num_layers=1,
has_bias=True,
batch_first=False,
dropout=0.0,
bidirectional=False):
super(StackLSTM, self).__init__()
self.num_layers = num_layers
self.batch_first = batch_first
self.transpose = P.Transpose()
# direction number
num_directions = 2 if bidirectional else 1
# input_size list
input_size_list = [input_size]
for i in range(num_layers - 1):
input_size_list.append(hidden_size * num_directions)
# layers
layers = []
for i in range(num_layers):
layers.append(
nn.LSTMCell(input_size=input_size_list[i],
hidden_size=hidden_size,
has_bias=has_bias,
batch_first=batch_first,
bidirectional=bidirectional,
dropout=dropout))
# weights
weights = []
for i in range(num_layers):
# weight size
weight_size = (input_size_list[i] +
hidden_size) * num_directions * hidden_size * 4
if has_bias:
bias_size = num_directions * hidden_size * 4
weight_size = weight_size + bias_size
# numpy weight
stdv = 1 / math.sqrt(hidden_size)
w_np = np.random.uniform(-stdv, stdv,
(weight_size, 1, 1)).astype(np.float32)
# lstm weight
weights.append(
Parameter(initializer(Tensor(w_np), w_np.shape),
name="weight" + str(i)))
#
self.lstms = layers
self.weight = ParameterTuple(tuple(weights))
def __init__(self,
input_size,
hidden_size,
num_layers=1,
has_bias=True,
batch_first=False,
dropout=0,
bidirectional=False):
super(LSTM, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.num_layers = num_layers
self.has_bias = has_bias
self.batch_first = validator.check_value_type("batch_first",
batch_first, [bool],
self.cls_name)
self.hidden_size = validator.check_integer("hidden_size", hidden_size,
0, Rel.GT, self.cls_name)
self.num_layers = validator.check_integer("num_layers", num_layers, 0,
Rel.GT, self.cls_name)
self.dropout = float(dropout)
self.bidirectional = bidirectional
if self.batch_first:
self.transpose1 = P.Transpose()
self.transpose2 = P.Transpose()
num_directions = 2 if self.bidirectional else 1
self.cpu_target = False
enable_debug = context.get_context("enable_debug_runtime")
if context.get_context("device_target") == "CPU" and not enable_debug:
self.cpu_target = True
if not self.cpu_target:
self.lstm = P.LSTM(input_size=self.input_size,
hidden_size=self.hidden_size,
num_layers=self.num_layers,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout)
weight_size = 0
gate_size = 4 * self.hidden_size
for layer in range(self.num_layers):
input_layer_size = self.input_size if layer == 0 else self.hidden_size * num_directions
increment_size = gate_size * input_layer_size
increment_size += gate_size * self.hidden_size
if self.has_bias:
increment_size += 2 * gate_size
weight_size += increment_size * num_directions
stdv = 1 / math.sqrt(hidden_size)
w_np = np.random.uniform(-stdv, stdv,
(weight_size, 1, 1)).astype(np.float32)
self.weight = Parameter(initializer(Tensor(w_np),
[weight_size, 1, 1]),
name='weight')
else:
input_size_list = []
input_size_list.append(self.input_size)
for i in range(self.num_layers - 1):
input_size_list.append(self.hidden_size * num_directions)
weights = []
layers = []
bias_size = 0 if not self.has_bias else num_directions * self.hidden_size * 4
stdv = 1 / math.sqrt(hidden_size)
for i in range(num_layers):
weight_size = (input_size_list[i] + self.hidden_size
) * num_directions * self.hidden_size * 4
if has_bias:
weight_size = weight_size + bias_size
w_np = np.random.uniform(
-stdv, stdv, (weight_size, 1, 1)).astype(np.float32)
weights.append(
Parameter(initializer(Tensor(w_np), w_np.shape),
name='weight' + str(i)))
layers.append(
nn.LSTMCell(input_size=input_size_list[i],
hidden_size=self.hidden_size,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout))
self.lstms = layers
self.weight = ParameterTuple(tuple(weights))
self.fill = P.Fill()
self.shape = P.Shape()
def __init__(self,
input_size,
hidden_size,
num_layers=1,
has_bias=True,
batch_first=False,
dropout=0,
bidirectional=False):
super(LSTM, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.num_layers = num_layers
self.has_bias = has_bias
self.batch_first = validator.check_value_type("batch_first",
batch_first, [bool],
self.cls_name)
self.dropout = float(dropout)
self.bidirectional = bidirectional
if self.batch_first:
self.transpose1 = P.Transpose()
self.transpose2 = P.Transpose()
num_directions = 2 if self.bidirectional else 1
self.cpu_target = False
if context.get_context("device_target") == "CPU":
self.cpu_target = True
if not self.cpu_target:
self.lstm = P.LSTM(input_size=self.input_size,
hidden_size=self.hidden_size,
num_layers=self.num_layers,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout)
weight_size = 0
gate_size = 4 * self.hidden_size
for layer in range(self.num_layers):
input_layer_size = self.input_size if layer == 0 else self.hidden_size * num_directions
increment_size = gate_size * input_layer_size
increment_size += gate_size * self.hidden_size
if self.has_bias:
increment_size += 2 * gate_size
weight_size += increment_size * num_directions
self.weight = Parameter(initializer(0.0, [weight_size, 1, 1]),
name='weight')
else:
layer = []
layer.append(
nn.LSTMCell(input_size=self.input_size,
hidden_size=self.hidden_size,
layer_index=0,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout))
for i in range(num_layers - 1):
layer.append(
nn.LSTMCell(input_size=self.hidden_size * num_directions,
hidden_size=self.hidden_size,
layer_index=i + 1,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout))
self.lstms = layer
self.fill = P.Fill()
self.shape = P.Shape()
def __init__(self,
batch_size,
input_size,
hidden_size,
num_layers,
bidirectional=False,
batch_norm=False,
rnn_type='LSTM'):
super(BatchRNN, self).__init__()
self.batch_size = batch_size
self.input_size = input_size
self.hidden_size = hidden_size
self.num_layers = num_layers
self.rnn_type = rnn_type
self.bidirectional = bidirectional
self.has_bias = True
self.is_batch_norm = batch_norm
self.num_directions = 2 if bidirectional else 1
self.reshape_op = P.Reshape()
self.shape_op = P.Shape()
self.sum_op = P.ReduceSum()
input_size_list = [input_size]
for i in range(num_layers - 1):
input_size_list.append(hidden_size)
layers = []
for i in range(num_layers):
layers.append(
nn.LSTMCell(input_size=input_size_list[i],
hidden_size=hidden_size,
bidirectional=bidirectional,
has_bias=self.has_bias))
weights = []
for i in range(num_layers):
weight_size = (input_size_list[i] +
hidden_size) * hidden_size * self.num_directions * 4
if self.has_bias:
bias_size = self.num_directions * hidden_size * 4 * 2
weight_size = weight_size + bias_size
stdv = 1 / math.sqrt(hidden_size)
w_np = np.random.uniform(-stdv, stdv,
(weight_size, 1, 1)).astype(np.float32)
weights.append(
Parameter(initializer(Tensor(w_np), w_np.shape),
name="weight" + str(i)))
self.h, self.c = self.stack_lstm_default_state(
batch_size,
hidden_size,
num_layers=num_layers,
bidirectional=bidirectional)
self.lstms = layers
self.weight = ParameterTuple(tuple(weights))
if batch_norm:
batch_norm_layer = []
for i in range(num_layers - 1):
batch_norm_layer.append(nn.BatchNorm1d(hidden_size))
self.batch_norm_list = batch_norm_layer
def __init__(self,
input_size,
hidden_size,
num_layers=1,
has_bias=True,
batch_first=False,
dropout=0,
bidirectional=False):
super(LSTM, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.num_layers = num_layers
self.has_bias = has_bias
self.batch_first = validator.check_value_type("batch_first", batch_first, [bool], self.cls_name)
self.dropout = float(dropout)
self.bidirectional = bidirectional
if self.batch_first:
self.transpose1 = P.Transpose()
self.transpose2 = P.Transpose()
num_directions = 2 if self.bidirectional else 1
self.cpu_target = False
if context.get_context("device_target") == "CPU":
self.cpu_target = True
if not self.cpu_target:
self.lstm = P.LSTM(input_size=self.input_size,
hidden_size=self.hidden_size,
num_layers=self.num_layers,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout)
weight_size = 0
gate_size = 4 * self.hidden_size
for layer in range(self.num_layers):
input_layer_size = self.input_size if layer == 0 else self.hidden_size * num_directions
increment_size = gate_size * input_layer_size
increment_size += gate_size * self.hidden_size
if self.has_bias:
increment_size += 2 * gate_size
weight_size += increment_size * num_directions
self.weight = Parameter(initializer(0.0, [weight_size, 1, 1]), name='weight')
else:
input_size_list = []
input_size_list.append(self.input_size)
for i in range(self.num_layers - 1):
input_size_list.append(self.hidden_size * num_directions)
weights = []
layers = []
bias_size = 0 if not self.has_bias else num_directions * self.hidden_size * 4
for i in range(num_layers):
weight_size = (input_size_list[i] + self.hidden_size) * num_directions * self.hidden_size * 4
w_np = np.ones([weight_size, 1, 1]).astype(np.float32) * 0.01
if has_bias:
bias_np = np.zeros([bias_size, 1, 1]).astype(np.float32)
w_np = np.concatenate([w_np, bias_np], axis=0)
weights.append(Parameter(initializer(Tensor(w_np), w_np.shape), name='weight' + str(i)))
layers.append(nn.LSTMCell(input_size=input_size_list[i],
hidden_size=self.hidden_size,
has_bias=self.has_bias,
bidirectional=self.bidirectional,
dropout=self.dropout))
self.lstms = layers
self.weight = ParameterTuple(tuple(weights))
self.fill = P.Fill()
self.shape = P.Shape()