Newer
Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
# ////////////////////////////////////////////////////////////////////////////////////////////////////////////
# /////////////////////////// Custom batchnorm implementing actual hardware ABN //////////////////////////////
# ////////////////////////////////////////////////////////////////////////////////////////////////////////////
# Inspired from https://stackoverflow.com/questions/54101593/conditional-batch-normalization-in-keras
import numpy as np
import math
import tensorflow as tf
import keras.backend as K
from keras import regularizers, initializers, constraints
#from keras.legacy import interfaces
from keras.layers import Layer, Input, InputSpec
from keras.models import Model
# Current ABN model
from models.ABN_charge import makeLookupABN, doInterpABN
from models.ABN_charge import round_through, floor_through
class Analog_BN(Layer):
""" Analog batchnorm layer
"""
# /// Init layer ///
# @interfaces.legacy_batchnorm_support
def __init__(self,
axis=-1,
momentum=0.99,
epsilon=1e-5,
center=True,
scale=True,
renorm = True,
beta_initializer='zeros',
gamma_initializer=tf.keras.initializers.Constant(value=3),
moving_mean_initializer='zeros',
moving_variance_initializer='ones',
beta_regularizer=None,
gamma_regularizer=None,
activity_regularizer=None,
beta_constraint=None,
gamma_constraint=None,
hardware = None,
NB = None,
m_sigma = 1,
Npoints = 401,
EN_NOISE = 0,
**kwargs):
super(Analog_BN, self).__init__(**kwargs)
self.axis = axis
self.momentum = momentum
self.epsilon = epsilon
self.center = center
self.scale = scale
self.renorm = renorm
self.beta_initializer = initializers.get(beta_initializer)
self.gamma_initializer = initializers.get(gamma_initializer)
self.moving_mean_initializer = initializers.get(moving_mean_initializer)
self.moving_variance_initializer = (initializers.get(moving_variance_initializer))
self.beta_regularizer = regularizers.get(beta_regularizer)
self.gamma_regularizer = regularizers.get(gamma_regularizer)
self.activity_regularizer = regularizers.get(activity_regularizer)
self.beta_constraint = constraints.get(beta_constraint)
self.gamma_constraint = constraints.get(gamma_constraint)
self.hardware = hardware;
self.EN_NOISE = EN_NOISE;
self.m_sigma = m_sigma;
# -- Interpolation info --
self.Npoints = Npoints;
self.ABN_lookup = None;
self.sig_ABN_lookup = None;
# /// Build layer ///
def build(self,input_shape):
dim = input_shape[self.axis];
if dim is None:
raise ValueError('Axis ' + str(self.axis) + ' of '
'input tensor should have a defined dimension '
'but the layer received an input with shape ' +
str(input_shape) + '.')
shape = (dim,)
if self.scale:
# gamma_constraint = Clip(0.0,4.0)
self.gamma = self.add_weight(shape = (1,),
name = 'gamma',
initializer = self.gamma_initializer,
regularizer = self.gamma_regularizer,
constraint = self.gamma_constraint)
else:
self.gamma = None
if self.center:
# beta_constraint = Clip(-100.0,100.0);
self.beta = self.add_weight(shape = shape,
name = 'beta',
initializer = self.beta_initializer,
regularizer = self.beta_regularizer,
constraint = self.beta_constraint)
else:
self.beta = None
if self.renorm:
self.moving_mean_DP = self.add_weight(
shape=shape,
name='moving_mean_DP',
initializer=self.moving_mean_initializer,
trainable=False)
self.moving_variance_DP = self.add_weight(
shape=shape,
name='moving_variance_DP',
initializer=self.moving_variance_initializer,
trainable=False)
else:
self.moving_mean_DP = K.variable(0.0)
self.moving_variance_DP = K.variable(1.0)
self.m_sigma = self.add_weight(shape = (1,),
name = 'm_sigma',
initializer = initializers.get(tf.keras.initializers.Constant(value=self.m_sigma)),
trainable=False)
# Spice-extracted lookup table between D_OUT, V_DP and T_ABN (hardcoded hardware has to match CIM params)
print('Retrieving actual charge-domain ABN response...')
self.ABN_lookup = self.hardware.sramInfo.ABN_LUT;
self.sig_ABN_lookup = self.hardware.sramInfo.sig_ABN_LUT;
print('Done !')
super(Analog_BN, self).build(input_shape)
# /// Call layer (train or inference) ///
def call(self,inputs,training=None):
input_shape = K.int_shape(inputs);
# print(input_shape)
# Prepare broadcasting shape.
ndim = len(input_shape)
reduction_axes = list(range(len(input_shape)))
del reduction_axes[self.axis]
broadcast_shape = [1] * len(input_shape)
broadcast_shape[self.axis] = input_shape[self.axis]
# Determines whether broadcasting is needed.
needs_broadcasting = (sorted(reduction_axes) != list(range(ndim))[:-1])
def normalize_inference():
# Explicitely broadcast parameters when required.
if needs_broadcasting:
# Norm params
if self.renorm:
broadcast_moving_mean_DP = K.reshape(self.moving_mean_DP,
broadcast_shape);
broadcast_moving_variance_DP = K.reshape(self.moving_variance_DP,
broadcast_shape);
else:
broadcast_moving_mean_DP = None;
broadcast_moving_variance_DP = None;
# Scale param
if self.scale:
broadcast_gamma = K.reshape(self.gamma,broadcast_shape);
else:
broadcast_gamma = None
# Offset param
if self.center:
broadcast_beta = K.reshape(self.beta,broadcast_shape);
else:
broadcast_beta = None
# Return batchnorm
return ABN(
inputs,
self.ABN_lookup,
self.sig_ABN_lookup,
self.V_DP_half_LUT,
self.devGainLUT,
broadcast_moving_mean_DP,
broadcast_moving_variance_DP,
broadcast_beta,
broadcast_gamma,
axis = self.axis,
epsilon = self.epsilon,
m_sigma = self.m_sigma,
hardware = self.hardware,
Npoints = self.Npoints,
EN_NOISE=self.EN_NOISE,
training=training)
else:
return ABN(
inputs,
self.ABN_lookup,
self.sig_ABN_lookup,
self.V_DP_half_LUT,
self.devGainLUT,
self.moving_mean_DP,
self.moving_variance_DP,
self.beta,
self.gamma,
axis = self.axis,
epsilon = self.epsilon,
m_sigma = self.m_sigma,
hardware = self.hardware,
Npoints = self.Npoints,
EN_NOISE=self.EN_NOISE,
training=training)
# If the learning phase is *static* and set to inference:
if training in {0, False}:
return normalize_inference()
# If the learning is either dynamic, or set to training:
(normed_training,mean_DP,variance_DP) = \
norm_ABN_in_train(
inputs,self.ABN_lookup,self.sig_ABN_lookup,self.V_DP_half_LUT, self.devGainLUT, self.beta, self.gamma, self.renorm, reduction_axes,
epsilon=self.epsilon,m_sigma=self.m_sigma,hardware=self.hardware,Npoints=self.Npoints,EN_NOISE=self.EN_NOISE,training=training)
# ???
if K.backend() != 'cntk':
sample_size = K.prod([K.shape(inputs)[axis]
for axis in reduction_axes])
sample_size = K.cast(sample_size, dtype=K.dtype(inputs))
if K.backend() == 'tensorflow' and sample_size.dtype != 'float32':
sample_size = K.cast(sample_size, dtype='float32')
# sample variance - unbiased estimator of population variance
variance_DP *= sample_size / (sample_size - (1.0 + self.epsilon))
# Update moving mean and variance during training
self.add_update([K.moving_average_update(self.moving_mean_DP,
mean_DP,
self.momentum),
K.moving_average_update(self.moving_variance_DP,
variance_DP,
self.momentum)])
# Pick ABN result for either training or inference
return K.in_train_phase(normed_training,
normalize_inference,
training=training)
def get_config(self):
config = {
'axis': self.axis,
'momentum': self.momentum,
'epsilon': self.epsilon,
'center': self.center,
'scale': self.scale,
'renorm': self.renorm,
'beta_initializer': initializers.serialize(self.beta_initializer),
'gamma_initializer': initializers.serialize(self.gamma_initializer),
'moving_mean_initializer':
initializers.serialize(self.moving_mean_initializer),
'moving_variance_initializer':
initializers.serialize(self.moving_variance_initializer),
'beta_regularizer': regularizers.serialize(self.beta_regularizer),
'gamma_regularizer': regularizers.serialize(self.gamma_regularizer),
'beta_constraint': constraints.serialize(self.beta_constraint),
'gamma_constraint': constraints.serialize(self.gamma_constraint)
}
base_config = super(Analog_BN, self).get_config()
return dict(list(base_config.items()) + list(config.items()))
def compute_output_shape(self, input_shape):
return input_shape[1]
############################################## Internal functions ##################################################
# Perform ABN
def ABN(V_DP,ABN_lookup,sig_ABN_lookup,V_DP_half_LUT,devGainLUT,mov_mean_DP=0.0,mov_variance_DP=1.0, beta=0.0,gamma=0.0,axis=-1,epsilon=1e-5,m_sigma=1,hardware=None,Npoints=401,EN_NOISE=0,training=False):
# Get hardware parameters
VDD = hardware.sramInfo.VDD.data;
r_gamma = hardware.sramInfo.r_gamma;
r_beta = hardware.sramInfo.r_beta;
OAres = hardware.sramInfo.OAres;
# Get number of states
Ns_gamma = 2**r_gamma;
Vmax_beta = hardware.sramInfo.Vmax_beta;
Vlsb_beta = Vmax_beta/2**(r_beta-1);
Vadc_step = VDD/(2**OAres);
# Set 'None' parameters to their initial values
if gamma is None:
gamma = K.constant(1.0);
if beta is None:
beta = K.constant(0.0);
if mov_mean_DP is None:
mov_mean_DP = K.constant(DR_tuple[1]);
if mov_variance_DP is None:
mov_variance_DP = K.constant(1.0);
# Specify non-centernormalized correction factors
# mu_goal = VDD/2;
sigma_goal = VDD/m_sigma; var_goal = sigma_goal*sigma_goal;
# # Get custom renorm factors
# sigma_DP = K.sqrt(mov_variance_DP);
# mov_mean_DP_t = mov_mean_DP - mu_goal/sigma_goal*sigma_DP;
# # mov_mean_DP_t = K.zeros_like(mov_mean_DP);
mov_variance_DP_t = K.mean(mov_variance_DP)/var_goal;
# mov_variance_DP_t = mov_variance_DP/var_goal;
# # Get equivalent coefficients
# sigma_DP_t = K.sqrt(mov_variance_DP_t);
gamma_eq = gamma/(K.sqrt(mov_variance_DP_t) + epsilon);
beta_eq = beta/gamma_eq - mov_mean_DP;
# Restrict gain factor to power-of-2
log_gamma_eq = round_through(tf.math.log(gamma_eq)/tf.math.log(2.));
gamma_eq = K.pow(2.,log_gamma_eq);
# Quantize results
gamma_eq = K.clip(round_through(gamma_eq),1,2**r_gamma);
V_beta = K.clip(round_through(beta_eq/Vlsb_beta)*Vlsb_beta,-Vmax_beta,Vmax_beta);
# Apply quantized offset
V_ABN_temp = V_DP+V_beta;
# // Get ABN distribution from LUTs based on the gain/offset mapping //
D_OUT = doInterpABN(ABN_lookup,gamma_eq,V_ABN_temp,Ns_gamma,Ns_gamma,VDD,Npoints);
sig_D_OUT = doInterpABN(sig_ABN_lookup,gamma_eq,V_ABN_temp,Ns_gamma,Ns_gamma,VDD,Npoints);
sig_D_OUT = sig_D_OUT*K.random_normal(shape=tf.shape(D_OUT),mean=0.,stddev=1.,dtype='float32');
D_OUT = D_OUT + sig_D_OUT;
# Reshape into the right order
return D_OUT;
# Compute mean and variance of the batch then perform ABN with it, when enabled
def norm_ABN_in_train(V_DP,ABN_lookup,sig_ABN_lookup,V_DP_half_LUT,devGainLUT,beta=0.0,gamma=1.0,renorm=True,axis=-1,epsilon=1e-5,m_sigma=1,hardware=None,Npoints=401,EN_NOISE=0,training=False):
# Compute mean and variance of each batch when desired
if(renorm):
# Eventually reshape V_DP in case of CONV2D operation
Ncols = K.int_shape(V_DP)[-1];
V_DP_flat = tf.reshape(V_DP,(-1,Ncols));
# Get mean and variance
mean_DP = K.mean(V_DP_flat,axis=0);
variance_DP = K.var(V_DP_flat,axis=0);
else:
mean_DP = K.constant(0.0);
variance_DP = K.constant(1.0);
# Compute ABN with specified mean and variance
V_DP_BN = ABN(V_DP,ABN_lookup,sig_ABN_lookup,V_DP_half_LUT,devGainLUT,mean_DP,variance_DP,beta,gamma,axis,epsilon,m_sigma,hardware,Npoints,EN_NOISE,training);
# Return a tuple of BN_result, mean and variance
return (V_DP_BN,mean_DP,variance_DP);