125 lines
4.2 KiB
Python
125 lines
4.2 KiB
Python
#! /usr/bin/python
|
|
# -*- coding: utf-8 -*-
|
|
|
|
import tensorlayer as tl
|
|
from tensorlayer import logging
|
|
from tensorlayer.layers.core import Module
|
|
from tensorlayer.layers.utils import cabs, quantize_active, quantize_weight
|
|
|
|
__all__ = [
|
|
'DorefaDense',
|
|
]
|
|
|
|
|
|
class DorefaDense(Module):
|
|
"""The :class:`DorefaDense` class is a binary fully connected layer, which weights are 'bitW' bits and the output of the previous layer
|
|
are 'bitA' bits while inferencing.
|
|
|
|
Note that, the bias vector would not be binarized.
|
|
|
|
Parameters
|
|
----------
|
|
bitW : int
|
|
The bits of this layer's parameter
|
|
bitA : int
|
|
The bits of the output of previous layer
|
|
n_units : int
|
|
The number of units of this layer.
|
|
act : activation function
|
|
The activation function of this layer, usually set to ``tf.act.sign`` or apply :class:`Sign` after :class:`BatchNorm`.
|
|
use_gemm : boolean
|
|
If True, use gemm instead of ``tf.matmul`` for inferencing. (TODO).
|
|
W_init : initializer
|
|
The initializer for the weight matrix.
|
|
b_init : initializer or None
|
|
The initializer for the bias vector. If None, skip biases.
|
|
in_channels: int
|
|
The number of channels of the previous layer.
|
|
If None, it will be automatically detected when the layer is forwarded for the first time.
|
|
name : a str
|
|
A unique layer name.
|
|
|
|
Examples
|
|
--------
|
|
>>> net = tl.layers.Input([10, 784], name='input')
|
|
>>> net = tl.layers.DorefaDense(n_units=800, act=tl.ReLU, name='relu1')(net)
|
|
>>> output shape :(10, 800)
|
|
>>> net = tl.layers.DorefaDense(n_units=10, name='output')(net)
|
|
>>> output shape :(10, 10)
|
|
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
bitW=1,
|
|
bitA=3,
|
|
n_units=100,
|
|
act=None,
|
|
use_gemm=False,
|
|
W_init=tl.initializers.truncated_normal(stddev=0.05),
|
|
b_init=tl.initializers.constant(value=0.0),
|
|
in_channels=None,
|
|
name=None, #'dorefa_dense',
|
|
):
|
|
super().__init__(name, act=act)
|
|
self.bitW = bitW
|
|
self.bitA = bitA
|
|
self.n_units = n_units
|
|
self.use_gemm = use_gemm
|
|
self.W_init = W_init
|
|
self.b_init = b_init
|
|
self.in_channels = in_channels
|
|
|
|
if self.in_channels is not None:
|
|
self.build((None, self.in_channels))
|
|
self._built = True
|
|
|
|
logging.info(
|
|
"DorefaDense %s: %d %s" %
|
|
(self.name, n_units, self.act.__name__ if self.act is not None else 'No Activation')
|
|
)
|
|
|
|
def __repr__(self):
|
|
actstr = self.act.__name__ if self.act is not None else 'No Activation'
|
|
s = ('{classname}(n_units={n_units}, ' + actstr)
|
|
s += ', bitW={bitW}, bitA={bitA}'
|
|
if self.in_channels is not None:
|
|
s += ', in_channels=\'{in_channels}\''
|
|
if self.name is not None:
|
|
s += ', name=\'{name}\''
|
|
s += ')'
|
|
return s.format(classname=self.__class__.__name__, **self.__dict__)
|
|
|
|
def build(self, inputs_shape):
|
|
if len(inputs_shape) != 2:
|
|
raise Exception("The input dimension must be rank 2, please reshape or flatten it")
|
|
|
|
if self.in_channels is None:
|
|
self.in_channels = inputs_shape[1]
|
|
|
|
if self.use_gemm:
|
|
raise Exception("TODO. The current version use tf.matmul for inferencing.")
|
|
|
|
n_in = inputs_shape[-1]
|
|
self.W = self._get_weights("weights", shape=(n_in, self.n_units), init=self.W_init)
|
|
if self.b_init is not None:
|
|
self.b = self._get_weights("biases", shape=(self.n_units), init=self.b_init)
|
|
self.bias_add = tl.ops.BiasAdd()
|
|
self.matmul = tl.ops.MatMul()
|
|
|
|
def forward(self, inputs):
|
|
if self._forward_state == False:
|
|
if self._built == False:
|
|
self.build(tl.get_tensor_shape(inputs))
|
|
self._built = True
|
|
self._forward_state = True
|
|
|
|
inputs = quantize_active(cabs(inputs), self.bitA)
|
|
W_ = quantize_weight(self.W, self.bitW)
|
|
outputs = self.matmul(inputs, W_)
|
|
if self.b_init is not None:
|
|
outputs = self.bias_add(outputs, self.b)
|
|
if self.act:
|
|
outputs = self.act(outputs)
|
|
return outputs
|