Source code for mxnet.gluon.rnn.rnn_cell

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.

# coding: utf-8
# pylint: disable=no-member, invalid-name, protected-access, no-self-use
# pylint: disable=too-many-branches, too-many-arguments, no-self-use
# pylint: disable=too-many-lines, arguments-differ
"""Definition of various recurrent neural network cells."""
__all__ = ['RecurrentCell', 'HybridRecurrentCell',
           'RNNCell', 'LSTMCell', 'GRUCell',
           'SequentialRNNCell', 'HybridSequentialRNNCell', 'DropoutCell',
           'ModifierCell', 'ZoneoutCell', 'ResidualCell',
           'BidirectionalCell', 'VariationalDropoutCell', 'LSTMPCell']

from ... import np, npx, cpu
from ...util import use_np
from ...base import string_types, numeric_types, _as_list
from ..block import Block, HybridBlock
from ..parameter import Parameter
from ..utils import _indent
from .. import tensor_types
from ..nn import LeakyReLU


def _cells_state_info(cells, batch_size):
    return sum([c().state_info(batch_size) for c in cells], [])

def _cells_begin_state(cells, **kwargs):
    return sum([c().begin_state(**kwargs) for c in cells], [])

def _get_begin_state(cell, begin_state, inputs, batch_size):
    if begin_state is None:
        device = inputs.device if isinstance(inputs, tensor_types) else inputs[0].device
        with device:
            begin_state = cell.begin_state(func=np.zeros, batch_size=batch_size)
    return begin_state

def _format_sequence(length, inputs, layout, merge, in_layout=None):
    assert inputs is not None, \
        "unroll(inputs=None) has been deprecated. " \
        "Please create input variables outside unroll."

    axis = layout.find('T')
    batch_axis = layout.find('N')
    batch_size = 0
    in_axis = in_layout.find('T') if in_layout is not None else axis
    if isinstance(inputs, np.ndarray):
        batch_size = inputs.shape[batch_axis]
        if merge is False:
            assert length is None or length == inputs.shape[in_axis]
            inputs = _as_list(npx.slice_channel(inputs, axis=in_axis,
                                                num_outputs=inputs.shape[in_axis],
                                                squeeze_axis=1))
    else:
        assert isinstance(inputs, (list, tuple)), \
            "Only support MXNet numpy ndarray or list of MXNet numpy ndarrays as inputs"
        assert length is None or len(inputs) == length
        batch_size = inputs[0].shape[0]
        if merge is True:
            inputs = np.stack(inputs, axis=axis)
            in_axis = axis

    if isinstance(inputs, np.ndarray) and axis != in_axis:
        inputs = np.swapaxes(inputs, axis, in_axis)

    return inputs, axis, batch_size

def _mask_sequence_variable_length(data, length, valid_length, time_axis, merge):
    assert valid_length is not None
    if not isinstance(data, tensor_types):
        data = np.stack(data, axis=time_axis)
    outputs = npx.sequence_mask(data, sequence_length=valid_length, use_sequence_length=True,
                                axis=time_axis)
    if not merge:
        outputs = _as_list(npx.slice_channel(outputs, num_outputs=length, axis=time_axis,
                                             squeeze_axis=True))
    return outputs

def _reverse_sequences(sequences, unroll_step, valid_length=None):
    if valid_length is None:
        reversed_sequences = list(reversed(sequences))
    else:
        reversed_sequences = npx.sequence_reverse(np.stack(sequences, axis=0),
                                                  sequence_length=valid_length,
                                                  use_sequence_length=True)
        if unroll_step > 1:
            reversed_sequences = npx.slice_channel(reversed_sequences, axis=0,
                                                   num_outputs=unroll_step, squeeze_axis=True)
        else:
            reversed_sequences = [reversed_sequences[0]]

    return reversed_sequences


[docs]@use_np class RecurrentCell(Block): """Abstract base class for RNN cells """ def __init__(self): super(RecurrentCell, self).__init__() self._modified = False self.reset()
[docs] def reset(self): """Reset before re-using the cell for another graph.""" self._init_counter = -1 self._counter = -1 for cell in self._children.values(): cell().reset()
[docs] def state_info(self, batch_size=0): """shape and layout information of states""" raise NotImplementedError()
[docs] def begin_state(self, batch_size=0, func=np.zeros, **kwargs): """Initial state for this cell. Parameters ---------- func : callable, default symbol.zeros Function for creating initial state. For Symbol API, func can be `symbol.zeros`, `symbol.uniform`, `symbol.var etc`. Use `symbol.var` if you want to directly feed input as states. For NDArray API, func can be `ndarray.zeros`, `ndarray.ones`, etc. batch_size: int, default 0 Only required for NDArray API. Size of the batch ('N' in layout) dimension of input. **kwargs : Additional keyword arguments passed to func. For example `mean`, `std`, `dtype`, etc. Returns ------- states : nested list of Symbol Starting states for the first RNN step. """ assert not self._modified, \ "After applying modifier cells (e.g. ZoneoutCell) the base " \ "cell cannot be called directly. Call the modifier cell instead." states = [] for info in self.state_info(batch_size): if info is not None: info.update(kwargs) else: info = kwargs state = func(shape=info.pop("shape", ()), device=info.pop("device", cpu()), dtype=info.pop("dtype", "float32")) states.append(state) return states
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): """Unrolls an RNN cell across time steps. Parameters ---------- length : int Number of steps to unroll. inputs : Symbol, list of Symbol, or None If `inputs` is a single Symbol (usually the output of Embedding symbol), it should have shape (batch_size, length, ...) if `layout` is 'NTC', or (length, batch_size, ...) if `layout` is 'TNC'. If `inputs` is a list of symbols (usually output of previous unroll), they should all have shape (batch_size, ...). begin_state : nested list of Symbol, optional Input states created by `begin_state()` or output state of another cell. Created from `begin_state()` if `None`. layout : str, optional `layout` of input symbol. Only used if inputs is a single Symbol. merge_outputs : bool, optional If `False`, returns outputs as a list of Symbols. If `True`, concatenates output across time steps and returns a single symbol with shape (batch_size, length, ...) if layout is 'NTC', or (length, batch_size, ...) if layout is 'TNC'. If `None`, output whatever is faster. valid_length : Symbol, NDArray or None `valid_length` specifies the length of the sequences in the batch without padding. This option is especially useful for building sequence-to-sequence models where the input and output sequences would potentially be padded. If `valid_length` is None, all sequences are assumed to have the same length. If `valid_length` is a Symbol or NDArray, it should have shape (batch_size,). The ith element will be the length of the ith sequence in the batch. The last valid state will be return and the padded outputs will be masked with 0. Note that `valid_length` must be smaller or equal to `length`. Returns ------- outputs : list of Symbol or Symbol Symbol (if `merge_outputs` is True) or list of Symbols (if `merge_outputs` is False) corresponding to the output from the RNN from this unrolling. states : list of Symbol The new state of this RNN after this unrolling. The type of this symbol is same as the output of `begin_state()`. """ # pylint: disable=too-many-locals self.reset() inputs, axis, batch_size = _format_sequence(length, inputs, layout, False) begin_state = _get_begin_state(self, begin_state, inputs, batch_size) states = begin_state outputs = [] all_states = [] for i in range(length): output, states = self(inputs[i], states) outputs.append(output) if valid_length is not None: all_states.append(states) if valid_length is not None: states = [npx.sequence_last(np.stack(ele_list, axis=0), sequence_length=valid_length, use_sequence_length=True, axis=0) for ele_list in zip(*all_states)] outputs = _mask_sequence_variable_length(outputs, length, valid_length, axis, True) outputs, _, _ = _format_sequence(length, outputs, layout, merge_outputs) return outputs, states
#pylint: disable=no-self-use def _get_activation(self, inputs, activation, **kwargs): """Get activation function. Convert if is string""" func = {'tanh': np.tanh, 'relu': npx.relu, 'sigmoid': npx.sigmoid, 'softsign': npx.softsign}.get(activation) if func: return func(inputs, **kwargs) elif isinstance(activation, string_types): return npx.activation(inputs, act_type=activation, **kwargs) elif isinstance(activation, LeakyReLU): return npx.leaky_relu(inputs, act_type='leaky', slope=activation._alpha, **kwargs) return activation(inputs, **kwargs)
[docs] def forward(self, inputs, states): """Unrolls the recurrent cell for one time step. Parameters ---------- inputs : sym.Variable Input symbol, 2D, of shape (batch_size * num_units). states : list of sym.Variable RNN state from previous step or the output of begin_state(). Returns ------- output : Symbol Symbol corresponding to the output from the RNN when unrolling for a single time step. states : list of Symbol The new state of this RNN after this unrolling. The type of this symbol is same as the output of `begin_state()`. This can be used as an input state to the next time step of this RNN. See Also -------- begin_state: This function can provide the states for the first time step. unroll: This function unrolls an RNN for a given number of (>=1) time steps. """ # pylint: disable= arguments-differ self._counter += 1 return super(RecurrentCell, self).forward(inputs, states)
[docs]@use_np class HybridRecurrentCell(RecurrentCell, HybridBlock): """HybridRecurrentCell supports hybridize.""" def __init__(self): super(HybridRecurrentCell, self).__init__()
[docs] def forward(self, x, *args, **kwargs): raise NotImplementedError
[docs]@use_np class RNNCell(HybridRecurrentCell): r"""Elman RNN recurrent neural network cell. Each call computes the following function: .. math:: h_t = \tanh(w_{ih} * x_t + b_{ih} + w_{hh} * h_{(t-1)} + b_{hh}) where :math:`h_t` is the hidden state at time `t`, and :math:`x_t` is the hidden state of the previous layer at time `t` or :math:`input_t` for the first layer. If nonlinearity='relu', then `ReLU` is used instead of `tanh`. Parameters ---------- hidden_size : int Number of units in output symbol activation : str or Symbol, default 'tanh' Type of activation function. i2h_weight_initializer : str or Initializer Initializer for the input weights matrix, used for the linear transformation of the inputs. h2h_weight_initializer : str or Initializer Initializer for the recurrent weights matrix, used for the linear transformation of the recurrent state. i2h_bias_initializer : str or Initializer, default 'zeros' Initializer for the bias vector. h2h_bias_initializer : str or Initializer, default 'zeros' Initializer for the bias vector. input_size: int, default 0 The number of expected features in the input x. If not specified, it will be inferred from input. Inputs: - **data**: input tensor with shape `(batch_size, input_size)`. - **states**: a list of one initial recurrent state tensor with shape `(batch_size, num_hidden)`. Outputs: - **out**: output tensor with shape `(batch_size, num_hidden)`. - **next_states**: a list of one output recurrent state tensor with the same shape as `states`. """ def __init__(self, hidden_size, activation='tanh', i2h_weight_initializer=None, h2h_weight_initializer=None, i2h_bias_initializer='zeros', h2h_bias_initializer='zeros', input_size=0): super(RNNCell, self).__init__() self._hidden_size = hidden_size self._activation = activation self._input_size = input_size self.i2h_weight = Parameter('i2h_weight', shape=(hidden_size, input_size), init=i2h_weight_initializer, allow_deferred_init=True) self.h2h_weight = Parameter('h2h_weight', shape=(hidden_size, hidden_size), init=h2h_weight_initializer, allow_deferred_init=True) self.i2h_bias = Parameter('i2h_bias', shape=(hidden_size,), init=i2h_bias_initializer, allow_deferred_init=True) self.h2h_bias = Parameter('h2h_bias', shape=(hidden_size,), init=h2h_bias_initializer, allow_deferred_init=True)
[docs] def state_info(self, batch_size=0): return [{'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}]
def _alias(self): return 'rnn' def __repr__(self): s = '{name}({mapping}' if hasattr(self, '_activation'): s += ', {_activation}' s += ')' shape = self.i2h_weight.shape mapping = '{0} -> {1}'.format(shape[1] if shape[1] else None, shape[0]) return s.format(name=self.__class__.__name__, mapping=mapping, **self.__dict__)
[docs] def forward(self, inputs, states): device = inputs.device i2h = npx.fully_connected(inputs, weight=self.i2h_weight.data(device), bias=self.i2h_bias.data(device), num_hidden=self._hidden_size, no_bias=False) h2h = npx.fully_connected(states[0].to_device(device), weight=self.h2h_weight.data(device), bias=self.h2h_bias.data(device), num_hidden=self._hidden_size, no_bias=False) i2h_plus_h2h = i2h + h2h output = self._get_activation(i2h_plus_h2h, self._activation) return output, [output]
[docs] def infer_shape(self, i, x, is_bidirect): if i == 0: self.i2h_weight.shape = (self._hidden_size, x.shape[x.ndim-1]) else: nh = self._hidden_size if is_bidirect: nh *= 2 self.i2h_weight.shape = (self._hidden_size, nh)
[docs]@use_np class LSTMCell(HybridRecurrentCell): r"""Long-Short Term Memory (LSTM) network cell. Each call computes the following function: .. math:: \begin{array}{ll} i_t = sigmoid(W_{ii} x_t + b_{ii} + W_{hi} h_{(t-1)} + b_{hi}) \\ f_t = sigmoid(W_{if} x_t + b_{if} + W_{hf} h_{(t-1)} + b_{hf}) \\ g_t = \tanh(W_{ig} x_t + b_{ig} + W_{hc} h_{(t-1)} + b_{hg}) \\ o_t = sigmoid(W_{io} x_t + b_{io} + W_{ho} h_{(t-1)} + b_{ho}) \\ c_t = f_t * c_{(t-1)} + i_t * g_t \\ h_t = o_t * \tanh(c_t) \end{array} where :math:`h_t` is the hidden state at time `t`, :math:`c_t` is the cell state at time `t`, :math:`x_t` is the hidden state of the previous layer at time `t` or :math:`input_t` for the first layer, and :math:`i_t`, :math:`f_t`, :math:`g_t`, :math:`o_t` are the input, forget, cell, and out gates, respectively. Parameters ---------- hidden_size : int Number of units in output symbol. i2h_weight_initializer : str or Initializer Initializer for the input weights matrix, used for the linear transformation of the inputs. h2h_weight_initializer : str or Initializer Initializer for the recurrent weights matrix, used for the linear transformation of the recurrent state. i2h_bias_initializer : str or Initializer, default 'zeros' Initializer for the bias vector. h2h_bias_initializer : str or Initializer, default 'zeros' Initializer for the bias vector. input_size: int, default 0 The number of expected features in the input x. If not specified, it will be inferred from input. activation : str, default 'tanh' Activation type to use. See nd/symbol Activation for supported types. recurrent_activation : str, default 'sigmoid' Activation type to use for the recurrent step. See nd/symbol Activation for supported types. Inputs: - **data**: input tensor with shape `(batch_size, input_size)`. - **states**: a list of two initial recurrent state tensors. Each has shape `(batch_size, num_hidden)`. Outputs: - **out**: output tensor with shape `(batch_size, num_hidden)`. - **next_states**: a list of two output recurrent state tensors. Each has the same shape as `states`. """ # pylint: disable=too-many-instance-attributes def __init__(self, hidden_size, i2h_weight_initializer=None, h2h_weight_initializer=None, i2h_bias_initializer='zeros', h2h_bias_initializer='zeros', input_size=0, activation='tanh', recurrent_activation='sigmoid'): super(LSTMCell, self).__init__() self._hidden_size = hidden_size self._input_size = input_size self.i2h_weight = Parameter('i2h_weight', shape=(4*hidden_size, input_size), init=i2h_weight_initializer, allow_deferred_init=True) self.h2h_weight = Parameter('h2h_weight', shape=(4*hidden_size, hidden_size), init=h2h_weight_initializer, allow_deferred_init=True) self.i2h_bias = Parameter('i2h_bias', shape=(4*hidden_size,), init=i2h_bias_initializer, allow_deferred_init=True) self.h2h_bias = Parameter('h2h_bias', shape=(4*hidden_size,), init=h2h_bias_initializer, allow_deferred_init=True) self._activation = activation self._recurrent_activation = recurrent_activation
[docs] def state_info(self, batch_size=0): return [{'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}, {'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}]
def _alias(self): return 'lstm' def __repr__(self): s = '{name}({mapping})' shape = self.i2h_weight.shape mapping = '{0} -> {1}'.format(shape[1] if shape[1] else None, shape[0]) return s.format(name=self.__class__.__name__, mapping=mapping, **self.__dict__)
[docs] def forward(self, inputs, states): # pylint: disable=too-many-locals device = inputs.device i2h = npx.fully_connected(inputs, weight=self.i2h_weight.data(device), bias=self.i2h_bias.data(device), num_hidden=self._hidden_size*4, no_bias=False) h2h = npx.fully_connected(states[0].to_device(device), weight=self.h2h_weight.data(device), bias=self.h2h_bias.data(device), num_hidden=self._hidden_size*4, no_bias=False) gates = i2h + h2h slice_gates = npx.slice_channel(gates, num_outputs=4) in_gate = self._get_activation(slice_gates[0], self._recurrent_activation) forget_gate = self._get_activation(slice_gates[1], self._recurrent_activation) in_transform = self._get_activation(slice_gates[2], self._activation) out_gate = self._get_activation(slice_gates[3], self._recurrent_activation) next_c = np.multiply(forget_gate, states[1].to_device(device)) + \ np.multiply(in_gate, in_transform) next_h = np.multiply(out_gate, npx.activation(next_c, act_type=self._activation)) return next_h, [next_h, next_c]
[docs] def infer_shape(self, i, x, is_bidirect): if i == 0: self.i2h_weight.shape = (4*self._hidden_size, x.shape[x.ndim-1]) else: nh = self._hidden_size if is_bidirect: nh *= 2 self.i2h_weight.shape = (4*self._hidden_size, nh)
[docs]@use_np class GRUCell(HybridRecurrentCell): r"""Gated Rectified Unit (GRU) network cell. Note: this is an implementation of the cuDNN version of GRUs (slight modification compared to Cho et al. 2014; the reset gate :math:`r_t` is applied after matrix multiplication). Each call computes the following function: .. math:: \begin{array}{ll} r_t = sigmoid(W_{ir} x_t + b_{ir} + W_{hr} h_{(t-1)} + b_{hr}) \\ i_t = sigmoid(W_{ii} x_t + b_{ii} + W_{hi} h_{(t-1)} + b_{hi}) \\ n_t = \tanh(W_{in} x_t + b_{in} + r_t * (W_{hn} h_{(t-1)} + b_{hn})) \\ h_t = (1 - i_t) * n_t + i_t * h_{(t-1)} \\ \end{array} where :math:`h_t` is the hidden state at time `t`, :math:`x_t` is the hidden state of the previous layer at time `t` or :math:`input_t` for the first layer, and :math:`r_t`, :math:`i_t`, :math:`n_t` are the reset, input, and new gates, respectively. Parameters ---------- hidden_size : int Number of units in output symbol. i2h_weight_initializer : str or Initializer Initializer for the input weights matrix, used for the linear transformation of the inputs. h2h_weight_initializer : str or Initializer Initializer for the recurrent weights matrix, used for the linear transformation of the recurrent state. i2h_bias_initializer : str or Initializer, default 'zeros' Initializer for the bias vector. h2h_bias_initializer : str or Initializer, default 'zeros' Initializer for the bias vector. input_size: int, default 0 The number of expected features in the input x. If not specified, it will be inferred from input. activation : str, default 'tanh' Activation type to use. See nd/symbol Activation for supported types. recurrent_activation : str, default 'sigmoid' Activation type to use for the recurrent step. See nd/symbol Activation for supported types. Inputs: - **data**: input tensor with shape `(batch_size, input_size)`. - **states**: a list of one initial recurrent state tensor with shape `(batch_size, num_hidden)`. Outputs: - **out**: output tensor with shape `(batch_size, num_hidden)`. - **next_states**: a list of one output recurrent state tensor with the same shape as `states`. """ def __init__(self, hidden_size, i2h_weight_initializer=None, h2h_weight_initializer=None, i2h_bias_initializer='zeros', h2h_bias_initializer='zeros', input_size=0, activation='tanh', recurrent_activation='sigmoid'): super(GRUCell, self).__init__() self._hidden_size = hidden_size self._input_size = input_size self.i2h_weight = Parameter('i2h_weight', shape=(3*hidden_size, input_size), init=i2h_weight_initializer, allow_deferred_init=True) self.h2h_weight = Parameter('h2h_weight', shape=(3*hidden_size, hidden_size), init=h2h_weight_initializer, allow_deferred_init=True) self.i2h_bias = Parameter('i2h_bias', shape=(3*hidden_size,), init=i2h_bias_initializer, allow_deferred_init=True) self.h2h_bias = Parameter('h2h_bias', shape=(3*hidden_size,), init=h2h_bias_initializer, allow_deferred_init=True) self._activation = activation self._recurrent_activation = recurrent_activation
[docs] def state_info(self, batch_size=0): return [{'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}]
def _alias(self): return 'gru' def __repr__(self): s = '{name}({mapping})' shape = self.i2h_weight.shape mapping = '{0} -> {1}'.format(shape[1] if shape[1] else None, shape[0]) return s.format(name=self.__class__.__name__, mapping=mapping, **self.__dict__)
[docs] def forward(self, inputs, states): # pylint: disable=too-many-locals device = inputs.device prev_state_h = states[0].to_device(device) i2h = npx.fully_connected(inputs, weight=self.i2h_weight.data(device), bias=self.i2h_bias.data(device), num_hidden=self._hidden_size * 3, no_bias=False) h2h = npx.fully_connected(prev_state_h, weight=self.h2h_weight.data(device), bias=self.h2h_bias.data(device), num_hidden=self._hidden_size * 3, no_bias=False) i2h_r, i2h_z, i2h = npx.slice_channel(i2h, num_outputs=3) h2h_r, h2h_z, h2h = npx.slice_channel(h2h, num_outputs=3) reset_gate = self._get_activation(i2h_r + h2h_r, self._recurrent_activation) update_gate = self._get_activation(i2h_z + h2h_z, self._recurrent_activation) next_h_tmp = self._get_activation(i2h + np.multiply(reset_gate, h2h), self._activation) ones = np.ones(update_gate.shape) next_h = np.multiply((ones - update_gate), next_h_tmp) + np.multiply(update_gate, prev_state_h) return next_h, [next_h]
[docs] def infer_shape(self, i, x, is_bidirect): if i == 0: self.i2h_weight.shape = (3*self._hidden_size, x.shape[x.ndim-1]) else: nh = self._hidden_size if is_bidirect: nh *= 2 self.i2h_weight.shape = (3*self._hidden_size, nh)
[docs]@use_np class SequentialRNNCell(RecurrentCell): """Sequentially stacking multiple RNN cells.""" def __init__(self): super(SequentialRNNCell, self).__init__() self._layers = [] def __repr__(self): s = '{name}(\n{modstr}\n)' return s.format(name=self.__class__.__name__, modstr='\n'.join(['({i}): {m}'.format(i=i, m=_indent(m().__repr__(), 2)) for i, m in self._children.items()]))
[docs] def add(self, cell): """Appends a cell into the stack. Parameters ---------- cell : RecurrentCell The cell to add. """ self._layers.append(cell) self.register_child(cell)
[docs] def state_info(self, batch_size=0): return _cells_state_info(self._children.values(), batch_size)
[docs] def begin_state(self, **kwargs): assert not self._modified, \ "After applying modifier cells (e.g. ZoneoutCell) the base " \ "cell cannot be called directly. Call the modifier cell instead." return _cells_begin_state(self._children.values(), **kwargs)
def __call__(self, inputs, states): self._counter += 1 next_states = [] p = 0 assert all(not isinstance(cell(), BidirectionalCell) for cell in self._children.values()) for cell in self._children.values(): assert not isinstance(cell(), BidirectionalCell) n = len(cell().state_info()) state = states[p:p+n] p += n inputs, state = cell()(inputs, state) next_states.append(state) return inputs, sum(next_states, [])
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): # pylint: disable=too-many-locals self.reset() inputs, _, batch_size = _format_sequence(length, inputs, layout, None) num_cells = len(self._children) begin_state = _get_begin_state(self, begin_state, inputs, batch_size) p = 0 next_states = [] for i, cell in enumerate(self._children.values()): n = len(cell().state_info()) states = begin_state[p:p+n] p += n inputs, states = cell().unroll(length, inputs=inputs, begin_state=states, layout=layout, merge_outputs=None if i < num_cells-1 else merge_outputs, valid_length=valid_length) next_states.extend(states) return inputs, next_states
def __getitem__(self, i): return self._children[str(i)]() def __len__(self): return len(self._children)
[docs] def forward(self, *args, **kwargs): # pylint: disable=missing-docstring raise NotImplementedError
def infer_shape(self, _, x, is_bidirect): for i, child in enumerate(self._layers): child.infer_shape(i, x, is_bidirect)
[docs]@use_np class HybridSequentialRNNCell(HybridRecurrentCell): """Sequentially stacking multiple HybridRNN cells.""" def __init__(self): super(HybridSequentialRNNCell, self).__init__() self._layers = [] def __repr__(self): s = '{name}(\n{modstr}\n)' return s.format(name=self.__class__.__name__, modstr='\n'.join(['({i}): {m}'.format(i=i, m=_indent(m().__repr__(), 2)) for i, m in self._children.items()]))
[docs] def add(self, cell): """Appends a cell into the stack. Parameters ---------- cell : RecurrentCell The cell to add. """ self._layers.append(cell) self.register_child(cell)
[docs] def state_info(self, batch_size=0): return _cells_state_info(self._children.values(), batch_size)
[docs] def begin_state(self, **kwargs): assert not self._modified, \ "After applying modifier cells (e.g. ZoneoutCell) the base " \ "cell cannot be called directly. Call the modifier cell instead." return _cells_begin_state(self._children.values(), **kwargs)
def __call__(self, inputs, states): self._counter += 1 next_states = [] p = 0 assert all(not isinstance(cell(), BidirectionalCell) for cell in self._children.values()) for cell in self._children.values(): n = len(cell().state_info()) state = states[p:p+n] p += n inputs, state = cell()(inputs, state) next_states.append(state) return inputs, sum(next_states, [])
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): self.reset() inputs, _, batch_size = _format_sequence(length, inputs, layout, None) num_cells = len(self._children) begin_state = _get_begin_state(self, begin_state, inputs, batch_size) p = 0 next_states = [] for i, cell in enumerate(self._children.values()): n = len(cell().state_info()) states = begin_state[p:p+n] p += n inputs, states = cell().unroll(length, inputs=inputs, begin_state=states, layout=layout, merge_outputs=None if i < num_cells-1 else merge_outputs, valid_length=valid_length) next_states.extend(states) return inputs, next_states
def __getitem__(self, i): return self._children[str(i)]() def __len__(self): return len(self._children)
[docs] def forward(self, inputs, states): return self.__call__(inputs, states)
# pylint: disable=unused-argument
[docs] def infer_shape(self, _, x, is_bidirect): for i, child in enumerate(self._layers): child.infer_shape(i, x, False)
[docs]@use_np class DropoutCell(HybridRecurrentCell): """Applies dropout on input. Parameters ---------- rate : float Percentage of elements to drop out, which is 1 - percentage to retain. axes : tuple of int, default () The axes on which dropout mask is shared. If empty, regular dropout is applied. Inputs: - **data**: input tensor with shape `(batch_size, size)`. - **states**: a list of recurrent state tensors. Outputs: - **out**: output tensor with shape `(batch_size, size)`. - **next_states**: returns input `states` directly. """ def __init__(self, rate, axes=()): super(DropoutCell, self).__init__() assert isinstance(rate, numeric_types), "rate must be a number" self._rate = rate self._axes = axes def __repr__(self): s = '{name}(rate={_rate}, axes={_axes})' return s.format(name=self.__class__.__name__, **self.__dict__)
[docs] def state_info(self, batch_size=0): return []
def _alias(self): return 'dropout'
[docs] def forward(self, inputs, states): if self._rate > 0: inputs = npx.dropout(data=inputs, p=self._rate, axes=self._axes) return inputs, states
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): self.reset() inputs, _, _ = _format_sequence(length, inputs, layout, merge_outputs) if isinstance(inputs, tensor_types): return self.forward(inputs, begin_state if begin_state else []) return super(DropoutCell, self).unroll( length, inputs, begin_state=begin_state, layout=layout, merge_outputs=merge_outputs, valid_length=None)
[docs]@use_np class ModifierCell(HybridRecurrentCell): """Base class for modifier cells. A modifier cell takes a base cell, apply modifications on it (e.g. Zoneout), and returns a new cell. After applying modifiers the base cell should no longer be called directly. The modifier cell should be used instead. """ def __init__(self, base_cell): assert not base_cell._modified, \ f"Cell {base_cell.name} is already modified. One cell cannot be modified twice" base_cell._modified = True super(ModifierCell, self).__init__() self.base_cell = base_cell @property def params(self): return self.base_cell.params
[docs] def state_info(self, batch_size=0): return self.base_cell.state_info(batch_size)
[docs] def begin_state(self, func=np.zeros, **kwargs): assert not self._modified, \ "After applying modifier cells (e.g. DropoutCell) the base " \ "cell cannot be called directly. Call the modifier cell instead." self.base_cell._modified = False begin = self.base_cell.begin_state(func=func, **kwargs) self.base_cell._modified = True return begin
[docs] def forward(self, inputs, states): raise NotImplementedError
def __repr__(self): s = '{name}({base_cell})' return s.format(name=self.__class__.__name__, **self.__dict__)
[docs]@use_np class ZoneoutCell(ModifierCell): """Applies Zoneout on base cell.""" def __init__(self, base_cell, zoneout_outputs=0., zoneout_states=0.): assert not isinstance(base_cell, BidirectionalCell), \ "BidirectionalCell doesn't support zoneout since it doesn't support step. " \ "Please add ZoneoutCell to the cells underneath instead." assert not isinstance(base_cell, SequentialRNNCell) or not base_cell._bidirectional, \ "Bidirectional SequentialRNNCell doesn't support zoneout. " \ "Please add ZoneoutCell to the cells underneath instead." super(ZoneoutCell, self).__init__(base_cell) self.zoneout_outputs = zoneout_outputs self.zoneout_states = zoneout_states self._prev_output = None def __repr__(self): s = '{name}(p_out={zoneout_outputs}, p_state={zoneout_states}, {base_cell})' return s.format(name=self.__class__.__name__, **self.__dict__) def _alias(self): return 'zoneout'
[docs] def reset(self): super(ZoneoutCell, self).reset() self._prev_output = None
[docs] def forward(self, inputs, states): device = inputs.device cell, p_outputs, p_states = self.base_cell, self.zoneout_outputs, self.zoneout_states next_output, next_states = cell(inputs, states) mask = (lambda p, like: npx.dropout(np.ones(like.shape), p=p)) prev_output = self._prev_output if prev_output is None: prev_output = np.zeros(next_output.shape) output = (np.where(mask(p_outputs, next_output), next_output, prev_output) if p_outputs != 0. else next_output) states = ([np.where(mask(p_states, new_s), new_s, old_s.to_device(device)) for new_s, old_s in zip(next_states, states)] if p_states != 0. else next_states) self._prev_output = output return output, states
[docs] def infer_shape(self, i, x, is_bidirect): self.base_cell.infer_shape(i, x, is_bidirect)
[docs]@use_np class ResidualCell(ModifierCell): """ Adds residual connection as described in Wu et al, 2016 (https://arxiv.org/abs/1609.08144). Output of the cell is output of the base cell plus input. """ def __init__(self, base_cell): # pylint: disable=useless-super-delegation super(ResidualCell, self).__init__(base_cell)
[docs] def forward(self, inputs, states): output, states = self.base_cell(inputs, states) output = output + inputs return output, states
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): self.reset() self.base_cell._modified = False outputs, states = self.base_cell.unroll(length, inputs=inputs, begin_state=begin_state, layout=layout, merge_outputs=merge_outputs, valid_length=valid_length) self.base_cell._modified = True merge_outputs = isinstance(outputs, tensor_types) if merge_outputs is None else \ merge_outputs inputs, axis, _ = _format_sequence(length, inputs, layout, merge_outputs) if valid_length is not None: # mask the padded inputs to zero inputs = _mask_sequence_variable_length(inputs, length, valid_length, axis, merge_outputs) if merge_outputs: outputs = outputs + inputs else: outputs = [i + j for i, j in zip(outputs, inputs)] return outputs, states
[docs] def infer_shape(self, i, x, is_bidirect): self.base_cell.infer_shape(i, x, is_bidirect)
[docs]@use_np class BidirectionalCell(HybridRecurrentCell): """Bidirectional RNN cell. Parameters ---------- l_cell : RecurrentCell Cell for forward unrolling r_cell : RecurrentCell Cell for backward unrolling """ def __init__(self, l_cell, r_cell): super(BidirectionalCell, self).__init__() self.l_cell = l_cell self.r_cell = r_cell def __call__(self, inputs, states): raise NotImplementedError("Bidirectional cannot be stepped. Please use unroll") def __repr__(self): s = '{name}(forward={l_cell}, backward={r_cell})' return s.format(name=self.__class__.__name__, l_cell=self._children['l_cell'](), r_cell=self._children['r_cell']())
[docs] def state_info(self, batch_size=0): return _cells_state_info(self._children.values(), batch_size)
[docs] def begin_state(self, **kwargs): assert not self._modified, \ "After applying modifier cells (e.g. DropoutCell) the base " \ "cell cannot be called directly. Call the modifier cell instead." return _cells_begin_state(self._children.values(), **kwargs)
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): # pylint: disable=too-many-locals self.reset() inputs, axis, batch_size = _format_sequence(length, inputs, layout, False) reversed_inputs = list(_reverse_sequences(inputs, length, valid_length)) begin_state = _get_begin_state(self, begin_state, inputs, batch_size) states = begin_state l_cell, r_cell = [c() for c in self._children.values()] l_outputs, l_states = l_cell.unroll(length, inputs=inputs, begin_state=states[:len(l_cell.state_info(batch_size))], layout=layout, merge_outputs=merge_outputs, valid_length=valid_length) r_outputs, r_states = r_cell.unroll(length, inputs=reversed_inputs, begin_state=states[len(l_cell.state_info(batch_size)):], layout=layout, merge_outputs=False, valid_length=valid_length) reversed_r_outputs = _reverse_sequences(r_outputs, length, valid_length) if merge_outputs is None: merge_outputs = isinstance(l_outputs, tensor_types) l_outputs, _, _ = _format_sequence(None, l_outputs, layout, merge_outputs) reversed_r_outputs, _, _ = _format_sequence(None, reversed_r_outputs, layout, merge_outputs) if merge_outputs: reversed_r_outputs = np.stack(reversed_r_outputs, axis=axis) outputs = np.concatenate([l_outputs, reversed_r_outputs], axis=2) else: outputs = [np.concatenate([l_o, r_o], axis=1) for i, (l_o, r_o) in enumerate(zip(l_outputs, reversed_r_outputs))] if valid_length is not None: outputs = _mask_sequence_variable_length(outputs, length, valid_length, axis, merge_outputs) states = l_states + r_states return outputs, states
#pylint: disable=W0613
[docs] def infer_shape(self, i, x, is_bidirect): l_cell, r_cell = [c() for c in self._children.values()] l_cell.infer_shape(i, x, True) r_cell.infer_shape(i, x, True)
[docs]@use_np class VariationalDropoutCell(ModifierCell): """ Applies Variational Dropout on base cell. https://arxiv.org/pdf/1512.05287.pdf Variational dropout uses the same dropout mask across time-steps. It can be applied to RNN inputs, outputs, and states. The masks for them are not shared. The dropout mask is initialized when stepping forward for the first time and will remain the same until .reset() is called. Thus, if using the cell and stepping manually without calling .unroll(), the .reset() should be called after each sequence. Parameters ---------- base_cell : RecurrentCell The cell on which to perform variational dropout. drop_inputs : float, default 0. The dropout rate for inputs. Won't apply dropout if it equals 0. drop_states : float, default 0. The dropout rate for state inputs on the first state channel. Won't apply dropout if it equals 0. drop_outputs : float, default 0. The dropout rate for outputs. Won't apply dropout if it equals 0. """ def __init__(self, base_cell, drop_inputs=0., drop_states=0., drop_outputs=0.): assert not drop_states or not isinstance(base_cell, BidirectionalCell), \ "BidirectionalCell doesn't support variational state dropout. " \ "Please add VariationalDropoutCell to the cells underneath instead." assert not drop_states \ or not isinstance(base_cell, SequentialRNNCell) or not base_cell._bidirectional, \ "Bidirectional SequentialRNNCell doesn't support variational state dropout. " \ "Please add VariationalDropoutCell to the cells underneath instead." super(VariationalDropoutCell, self).__init__(base_cell) self.drop_inputs = drop_inputs self.drop_states = drop_states self.drop_outputs = drop_outputs self.drop_inputs_mask = None self.drop_states_mask = None self.drop_outputs_mask = None def _alias(self): return 'vardrop'
[docs] def reset(self): super(VariationalDropoutCell, self).reset() self.drop_inputs_mask = None self.drop_states_mask = None self.drop_outputs_mask = None
def _initialize_input_masks(self, inputs, states): if self.drop_states and self.drop_states_mask is None: self.drop_states_mask = npx.dropout(np.ones(states[0].shape), p=self.drop_states) if self.drop_inputs and self.drop_inputs_mask is None: self.drop_inputs_mask = npx.dropout(np.ones(inputs.shape), p=self.drop_inputs) def _initialize_output_mask(self, output): if self.drop_outputs and self.drop_outputs_mask is None: self.drop_outputs_mask = npx.dropout(np.ones(output.shape), p=self.drop_outputs)
[docs] def forward(self, inputs, states): device = inputs.device cell = self.base_cell self._initialize_input_masks(inputs, states) if self.drop_states: states = list(states) # state dropout only needs to be applied on h, which is always the first state. states[0] = states[0].to_device(device) * self.drop_states_mask if self.drop_inputs: inputs = inputs * self.drop_inputs_mask next_output, next_states = cell(inputs, states) self._initialize_output_mask(next_output) if self.drop_outputs: next_output = next_output * self.drop_outputs_mask return next_output, next_states
def __repr__(self): s = '{name}(p_out = {drop_outputs}, p_state = {drop_states})' return s.format(name=self.__class__.__name__, **self.__dict__)
[docs] def unroll(self, length, inputs, begin_state=None, layout='NTC', merge_outputs=None, valid_length=None): """Unrolls an RNN cell across time steps. Parameters ---------- length : int Number of steps to unroll. inputs : Symbol, list of Symbol, or None If `inputs` is a single Symbol (usually the output of Embedding symbol), it should have shape (batch_size, length, ...) if `layout` is 'NTC', or (length, batch_size, ...) if `layout` is 'TNC'. If `inputs` is a list of symbols (usually output of previous unroll), they should all have shape (batch_size, ...). begin_state : nested list of Symbol, optional Input states created by `begin_state()` or output state of another cell. Created from `begin_state()` if `None`. layout : str, optional `layout` of input symbol. Only used if inputs is a single Symbol. merge_outputs : bool, optional If `False`, returns outputs as a list of Symbols. If `True`, concatenates output across time steps and returns a single symbol with shape (batch_size, length, ...) if layout is 'NTC', or (length, batch_size, ...) if layout is 'TNC'. If `None`, output whatever is faster. valid_length : Symbol, NDArray or None `valid_length` specifies the length of the sequences in the batch without padding. This option is especially useful for building sequence-to-sequence models where the input and output sequences would potentially be padded. If `valid_length` is None, all sequences are assumed to have the same length. If `valid_length` is a Symbol or NDArray, it should have shape (batch_size,). The ith element will be the length of the ith sequence in the batch. The last valid state will be return and the padded outputs will be masked with 0. Note that `valid_length` must be smaller or equal to `length`. Returns ------- outputs : list of Symbol or Symbol Symbol (if `merge_outputs` is True) or list of Symbols (if `merge_outputs` is False) corresponding to the output from the RNN from this unrolling. states : list of Symbol The new state of this RNN after this unrolling. The type of this symbol is same as the output of `begin_state()`. """ # Dropout on inputs and outputs can be performed on the whole sequence # only when state dropout is not present. if self.drop_states: return super(VariationalDropoutCell, self).unroll(length, inputs, begin_state, layout, merge_outputs, valid_length=valid_length) self.reset() inputs, axis, batch_size = _format_sequence(length, inputs, layout, True) states = _get_begin_state(self, begin_state, inputs, batch_size) if self.drop_inputs: inputs = npx.dropout(inputs, p=self.drop_inputs, axes=(axis,)) outputs, states = self.base_cell.unroll(length, inputs, states, layout, merge_outputs=True, valid_length=valid_length) if self.drop_outputs: outputs = npx.dropout(outputs, p=self.drop_outputs, axes=(axis,)) merge_outputs = isinstance(outputs, tensor_types) if merge_outputs is None else \ merge_outputs outputs, _, _ = _format_sequence(length, outputs, layout, merge_outputs) if valid_length is not None: outputs = _mask_sequence_variable_length(outputs, length, valid_length, axis, merge_outputs) return outputs, states
[docs] def infer_shape(self, i, x, is_bidirect): self.base_cell.infer_shape(i, x, is_bidirect)
[docs]@use_np class LSTMPCell(HybridRecurrentCell): r"""Long-Short Term Memory Projected (LSTMP) network cell. (https://arxiv.org/abs/1402.1128) Each call computes the following function: .. math:: \begin{array}{ll} i_t = sigmoid(W_{ii} x_t + b_{ii} + W_{ri} r_{(t-1)} + b_{ri}) \\ f_t = sigmoid(W_{if} x_t + b_{if} + W_{rf} r_{(t-1)} + b_{rf}) \\ g_t = \tanh(W_{ig} x_t + b_{ig} + W_{rc} r_{(t-1)} + b_{rg}) \\ o_t = sigmoid(W_{io} x_t + b_{io} + W_{ro} r_{(t-1)} + b_{ro}) \\ c_t = f_t * c_{(t-1)} + i_t * g_t \\ h_t = o_t * \tanh(c_t) \\ r_t = W_{hr} h_t \end{array} where :math:`r_t` is the projected recurrent activation at time `t`, :math:`h_t` is the hidden state at time `t`, :math:`c_t` is the cell state at time `t`, :math:`x_t` is the input at time `t`, and :math:`i_t`, :math:`f_t`, :math:`g_t`, :math:`o_t` are the input, forget, cell, and out gates, respectively. Parameters ---------- hidden_size : int Number of units in cell state symbol. projection_size : int Number of units in output symbol. i2h_weight_initializer : str or Initializer Initializer for the input weights matrix, used for the linear transformation of the inputs. h2h_weight_initializer : str or Initializer Initializer for the recurrent weights matrix, used for the linear transformation of the hidden state. h2r_weight_initializer : str or Initializer Initializer for the projection weights matrix, used for the linear transformation of the recurrent state. i2h_bias_initializer : str or Initializer, default 'lstmbias' Initializer for the bias vector. By default, bias for the forget gate is initialized to 1 while all other biases are initialized to zero. h2h_bias_initializer : str or Initializer Initializer for the bias vector. Inputs: - **data**: input tensor with shape `(batch_size, input_size)`. - **states**: a list of two initial recurrent state tensors, with shape `(batch_size, projection_size)` and `(batch_size, hidden_size)` respectively. Outputs: - **out**: output tensor with shape `(batch_size, num_hidden)`. - **next_states**: a list of two output recurrent state tensors. Each has the same shape as `states`. """ def __init__(self, hidden_size, projection_size, i2h_weight_initializer=None, h2h_weight_initializer=None, h2r_weight_initializer=None, i2h_bias_initializer='zeros', h2h_bias_initializer='zeros', input_size=0): super(LSTMPCell, self).__init__() self._hidden_size = hidden_size self._input_size = input_size self._projection_size = projection_size self.i2h_weight = Parameter('i2h_weight', shape=(4*hidden_size, input_size), init=i2h_weight_initializer, allow_deferred_init=True) self.h2h_weight = Parameter('h2h_weight', shape=(4*hidden_size, projection_size), init=h2h_weight_initializer, allow_deferred_init=True) self.h2r_weight = Parameter('h2r_weight', shape=(projection_size, hidden_size), init=h2r_weight_initializer, allow_deferred_init=True) self.i2h_bias = Parameter('i2h_bias', shape=(4*hidden_size,), init=i2h_bias_initializer, allow_deferred_init=True) self.h2h_bias = Parameter('h2h_bias', shape=(4*hidden_size,), init=h2h_bias_initializer, allow_deferred_init=True)
[docs] def state_info(self, batch_size=0): return [{'shape': (batch_size, self._projection_size), '__layout__': 'NC'}, {'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}]
def _alias(self): return 'lstmp' def __repr__(self): s = '{name}({mapping})' shape = self.i2h_weight.shape proj_shape = self.h2r_weight.shape mapping = '{0} -> {1} -> {2}'.format(shape[1] if shape[1] else None, shape[0], proj_shape[0]) return s.format(name=self.__class__.__name__, mapping=mapping, **self.__dict__) # pylint: disable= arguments-differ
[docs] def forward(self, inputs, states): device = inputs.device i2h = npx.fully_connected(inputs, weight=self.i2h_weight.data(device), bias=self.i2h_bias.data(device), num_hidden=self._hidden_size*4, no_bias=False) h2h = npx.fully_connected(states[0].to_device(device), weight=self.h2h_weight.data(device), bias=self.h2h_bias.data(device), num_hidden=self._hidden_size*4, no_bias=False) gates = i2h + h2h slice_gates = npx.slice_channel(gates, num_outputs=4) in_gate = npx.activation(slice_gates[0], act_type="sigmoid") forget_gate = npx.activation(slice_gates[1], act_type="sigmoid") in_transform = npx.activation(slice_gates[2], act_type="tanh") out_gate = npx.activation(slice_gates[3], act_type="sigmoid") next_c = forget_gate * states[1].to_device(device) + in_gate * in_transform hidden = np.multiply(out_gate, npx.activation(next_c, act_type="tanh")) next_r = npx.fully_connected(hidden, num_hidden=self._projection_size, weight=self.h2r_weight.data(device), no_bias=True) return next_r, [next_r, next_c]
[docs] def infer_shape(self, i, x, is_bidirect): if i == 0: self.i2h_weight.shape = (4*self._hidden_size, x.shape[x.ndim-1]) else: nh = self._projection_size if is_bidirect: nh *= 2 self.i2h_weight.shape = (4*self._hidden_size, nh)
def dynamic_unroll(cell, inputs, begin_state, drop_inputs=0, drop_outputs=0, layout='TNC', valid_length=None): """Unrolls an RNN cell across time steps. Currently, 'TNC' is a preferred layout. unroll on the input of this layout runs much faster. Parameters ---------- cell : an object whose base class is RNNCell. The RNN cell to run on the input sequence. inputs : Symbol It should have shape (batch_size, length, ...) if `layout` is 'NTC', or (length, batch_size, ...) if `layout` is 'TNC'. begin_state : nested list of Symbol The initial states of the RNN sequence. drop_inputs : float, default 0. The dropout rate for inputs. Won't apply dropout if it equals 0. drop_outputs : float, default 0. The dropout rate for outputs. Won't apply dropout if it equals 0. layout : str, optional `layout` of input symbol. Only used if inputs is a single Symbol. valid_length : Symbol, NDArray or None `valid_length` specifies the length of the sequences in the batch without padding. This option is especially useful for building sequence-to-sequence models where the input and output sequences would potentially be padded. If `valid_length` is None, all sequences are assumed to have the same length. If `valid_length` is a Symbol or NDArray, it should have shape (batch_size,). The ith element will be the length of the ith sequence in the batch. The last valid state will be return and the padded outputs will be masked with 0. Note that `valid_length` must be smaller or equal to `length`. Returns ------- outputs : Symbol the output of the RNN from this unrolling. states : list of Symbol The new state of this RNN after this unrolling. The type of this symbol is same as the output of `begin_state`. Examples -------- >>> seq_len = 3 >>> batch_size = 2 >>> input_size = 5 >>> cell = mx.gluon.rnn.LSTMCell(input_size) >>> cell.initialize(device=mx.cpu()) >>> rnn_data = mx.np.normal(loc=0, scale=1, shape=(seq_len, batch_size, input_size)) >>> state_shape = (batch_size, input_size) >>> states = [mx.np.normal(loc=0, scale=1, shape=state_shape) for i in range(2)] >>> valid_length = mx.np.array([2, 3]) >>> output, states = mx.gluon.rnn.rnn_cell.dynamic_unroll(cell, rnn_data, states, ... valid_length=valid_length, ... layout='TNC') >>> print(output) [[[ 0.00767238 0.00023103 0.03973929 -0.00925503 -0.05660512] [ 0.00881535 0.05428379 -0.02493718 -0.01834097 0.02189514]] [[-0.00676967 0.01447039 0.01287002 -0.00574152 -0.05734247] [ 0.01568508 0.02650866 -0.04270559 -0.04328435 0.00904011]] [[ 0. 0. 0. 0. 0. ] [ 0.01055336 0.02734251 -0.03153727 -0.03742751 -0.01378113]]] <NDArray 3x2x5 @cpu(0)> """ # Merge is always True, so we don't need length. inputs, axis, _ = _format_sequence(0, inputs, layout, True) if axis != 0: axes = list(range(len(layout))) tmp = axes[0] axes[0] = axes[axis] axes[axis] = tmp inputs = np.transpose(inputs, axes=axes) states = begin_state if drop_inputs: inputs = npx.dropout(inputs, p=drop_inputs, axes=(axis,)) if valid_length is None: outputs, states = npx.foreach(cell, inputs, states + [valid_length]) else: zeros = [] for s in states: zeros.append(np.zeros(s.shape)) states = list(_as_list(states)) states.append(np.zeros((1))) class loop_body(HybridBlock): """Loop body for foreach operator""" def __init__(self, cell): super(loop_body, self).__init__() self.cell = cell def forward(self, inputs, states): valid_len = states.pop() cell_states = states[:-1] iter_no = states[-1] out, new_states = self.cell(inputs, cell_states) for i, state in enumerate(cell_states): cond = npx.broadcast_greater(valid_len, iter_no) cond_broad = np.broadcast_to(cond, new_states[i].T.shape).T new_states[i] = np.where(cond_broad, new_states[i], state) new_states.append(iter_no + 1) new_states.append(valid_len) return out, new_states body = loop_body(cell) outputs, states = npx.foreach(body, inputs, states + [valid_length]) states.pop() if drop_outputs: outputs = npx.dropout(outputs, p=drop_outputs, axes=(axis,)) if valid_length is not None: if axis != 0: outputs = np.transpose(outputs, axes) outputs = npx.sequence_mask(outputs, sequence_length=valid_length, use_sequence_length=True, axis=axis) # the last state is the iteration number. We don't need it. return outputs, states[:-1] else: if axis != 0: outputs = np.transpose(outputs, axes) return outputs, states