ParakeetRebeccaRosario/parakeet/modules/customized.py

273 lines
10 KiB
Python

# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed 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.
from paddle import fluid
import paddle.fluid.layers as F
import paddle.fluid.dygraph as dg
class Pool1D(dg.Layer):
"""
A Pool 1D block implemented with Pool2D.
"""
def __init__(self,
pool_size=-1,
pool_type='max',
pool_stride=1,
pool_padding=0,
global_pooling=False,
use_cudnn=True,
ceil_mode=False,
exclusive=True,
data_format='NCT'):
super(Pool1D, self).__init__()
self.pool_size = pool_size
self.pool_type = pool_type
self.pool_stride = pool_stride
self.pool_padding = pool_padding
self.global_pooling = global_pooling
self.use_cudnn = use_cudnn
self.ceil_mode = ceil_mode
self.exclusive = exclusive
self.data_format = data_format
self.pool2d = dg.Pool2D(
[1, pool_size],
pool_type=pool_type,
pool_stride=[1, pool_stride],
pool_padding=[0, pool_padding],
global_pooling=global_pooling,
use_cudnn=use_cudnn,
ceil_mode=ceil_mode,
exclusive=exclusive)
def forward(self, x):
"""
Args:
x (Variable): Shape(B, C_in, 1, T), the input, where C_in means
input channels.
Returns:
x (Variable): Shape(B, C_out, 1, T), the outputs, where C_out means
output channels (num_filters).
"""
if self.data_format == 'NTC':
x = fluid.layers.transpose(x, [0, 2, 1])
x = fluid.layers.unsqueeze(x, [2])
x = self.pool2d(x)
x = fluid.layers.squeeze(x, [2])
if self.data_format == 'NTC':
x = fluid.layers.transpose(x, [0, 2, 1])
return x
class Conv1D(dg.Conv2D):
"""A standard Conv1D layer that use (B, C, T) data layout. It inherit Conv2D and
use (B, C, 1, T) data layout to compute 1D convolution. Nothing more.
NOTE: we inherit Conv2D instead of encapsulate a Conv2D layer to make it a simple
layer, instead of a complex one. So we can easily apply weight norm to it.
"""
def __init__(self,
num_channels,
num_filters,
filter_size,
stride=1,
padding=0,
dilation=1,
groups=1,
param_attr=None,
bias_attr=None,
use_cudnn=True,
act=None,
dtype='float32'):
super(Conv1D, self).__init__(
num_channels,
num_filters, (1, filter_size),
stride=(1, stride),
padding=(0, padding),
dilation=(1, dilation),
groups=groups,
param_attr=param_attr,
bias_attr=bias_attr,
use_cudnn=use_cudnn,
act=act,
dtype=dtype)
def forward(self, x):
"""Compute Conv1D by unsqueeze the input and squeeze the output.
Args:
x (Variable): shape(B, C_in, T_in), dtype float32, input of Conv1D.
Returns:
Variable: shape(B, C_out, T_out), dtype float32, output of Conv1D.
"""
x = F.unsqueeze(x, [2])
x = super(Conv1D, self).forward(x) # maybe risky here
x = F.squeeze(x, [2])
return x
class Conv1DTranspose(dg.Conv2DTranspose):
def __init__(self,
num_channels,
num_filters,
filter_size,
padding=0,
stride=1,
dilation=1,
groups=1,
param_attr=None,
bias_attr=None,
use_cudnn=True,
act=None,
dtype='float32'):
super(Conv1DTranspose, self).__init__(
num_channels,
num_filters, (1, filter_size),
output_size=None,
padding=(0, padding),
stride=(1, stride),
dilation=(1, dilation),
groups=groups,
param_attr=param_attr,
bias_attr=bias_attr,
use_cudnn=use_cudnn,
act=act,
dtype=dtype)
def forward(self, x):
"""Compute Conv1DTranspose by unsqueeze the input and squeeze the output.
Args:
x (Variable): shape(B, C_in, T_in), dtype float32, input of Conv1DTranspose.
Returns:
Variable: shape(B, C_out, T_out), dtype float32, output of Conv1DTranspose.
"""
x = F.unsqueeze(x, [2])
x = super(Conv1DTranspose, self).forward(x) # maybe risky here
x = F.squeeze(x, [2])
return x
class Conv1DCell(Conv1D):
"""A causal convolve-1d cell. It uses causal padding, padding(receptive_field -1, 0).
But Conv2D in dygraph does not support asymmetric padding yet, we just pad
(receptive_field -1, receptive_field -1) and drop last receptive_field -1 steps in
the output.
It is a cell that it acts like an RNN cell. It does not support stride > 1, and it
ensures 1-to-1 mapping from input time steps to output timesteps.
"""
def __init__(self,
num_channels,
num_filters,
filter_size,
dilation=1,
causal=False,
groups=1,
param_attr=None,
bias_attr=None,
use_cudnn=True,
act=None,
dtype='float32'):
receptive_field = 1 + dilation * (filter_size - 1)
padding = receptive_field - 1 if causal else receptive_field // 2
self._receptive_field = receptive_field
self.causal = causal
super(Conv1DCell, self).__init__(
num_channels,
num_filters,
filter_size,
stride=1,
padding=padding,
dilation=dilation,
groups=groups,
param_attr=param_attr,
bias_attr=bias_attr,
use_cudnn=use_cudnn,
act=act,
dtype=dtype)
def forward(self, x):
"""Compute Conv1D by unsqueeze the input and squeeze the output.
Args:
x (Variable): shape(B, C_in, T), dtype float32, input of Conv1D.
Returns:
Variable: shape(B, C_out, T), dtype float32, output of Conv1D.
"""
# it ensures that ouput time steps == input time steps
time_steps = x.shape[-1]
x = super(Conv1DCell, self).forward(x)
if x.shape[-1] != time_steps:
x = x[:, :, :time_steps]
return x
@property
def receptive_field(self):
return self._receptive_field
def start_sequence(self):
"""Prepare the Conv1DCell to generate a new sequence, this method should be called before calling add_input multiple times.
WARNING:
This method accesses `self.weight` directly. If a `Conv1DCell` object is wrapped in a `WeightNormWrapper`, make sure this method is called only after the `WeightNormWrapper`'s hook is called.
`WeightNormWrapper` removes the wrapped layer's `weight`, add has a `weight_v` and `weight_g` to re-compute the wrapped layer's weight as $weight = weight_g * weight_v / ||weight_v||$. (Recomputing the `weight` is a hook before calling the wrapped layer's `forward` method.)
Whenever a `WeightNormWrapper`'s `forward` method is called, the wrapped layer's weight is updated. But when loading from a checkpoint, `weight_v` and `weight_g` are updated but the wrapped layer's weight is not, since it is no longer a `Parameter`. You should manually call `remove_weight_norm` or `hook` to re-compute the wrapped layer's weight before calling this method if you don't call `forward` first.
So when loading a model which uses `Conv1DCell` objects wrapped in `WeightNormWrapper`s, remember to call `remove_weight_norm` for all `WeightNormWrapper`s before synthesizing. Also, removing weight norm speeds up computation.
"""
if not self.causal:
raise ValueError(
"Only causal conv1d shell should use start sequence")
if self.receptive_field == 1:
raise ValueError(
"Convolution block with receptive field = 1 does not need"
" to be implemented as a Conv1DCell. Conv1D suffices")
self._buffer = None
self._reshaped_weight = F.reshape(self.weight, (self._num_filters, -1))
def add_input(self, x_t):
"""This method works similarily with forward but in a `step-in-step-out` fashion.
Args:
x (Variable): shape(B, C_in, T=1), dtype float32, input of Conv1D.
Returns:
Variable: shape(B, C_out, T=1), dtype float32, output of Conv1D.
"""
batch_size, c_in, _ = x_t.shape
if self._buffer is None:
self._buffer = F.zeros(
(batch_size, c_in, self.receptive_field), dtype=x_t.dtype)
self._buffer = F.concat([self._buffer[:, :, 1:], x_t], -1)
if self._dilation[1] > 1:
input = F.strided_slice(
self._buffer,
axes=[2],
starts=[0],
ends=[self.receptive_field],
strides=[self._dilation[1]])
else:
input = self._buffer
input = F.reshape(input, (batch_size, -1))
y_t = F.matmul(input, self._reshaped_weight, transpose_y=True)
y_t = y_t + self.bias
y_t = F.unsqueeze(y_t, [-1])
return y_t