-
Notifications
You must be signed in to change notification settings - Fork 0
/
bit_flip.py
138 lines (115 loc) · 6.81 KB
/
bit_flip.py
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
import torch
import torch.nn as nn
import torch.nn.functional as F
from bin_int_convert import *
'''
Definition of function parameters:
@param wq_int: baseline quantized integer weight
@param w_bitwidth: bit width of the baseline quantized integer weight
@param group_size: group size for bit flip
@param num_pruned_column: number of desired zero bit-columns in every weight group
@param device: 'cpu' or 'cuda'
'''
def bitFlip_conv(wq_int, w_bitwidth: int=8, group_size: int=16, num_pruned_column: int=4, device='cpu'):
wq_int = wq_int.to(device)
K, C, H, W = wq_int.size() # output channel, input channel, kernel width, kernel height
if C < group_size:
group_size = C
NUM_GROUP = K*W*H*C//group_size
wq_int = wq_int.permute([0, 2, 3, 1]).unsqueeze(-1)
wq_int = wq_int.view(NUM_GROUP, group_size)
wqb_signMagnitude = int_to_signMagnitude(wq_int, w_bitwidth=w_bitwidth, device=device)
# check existing zero columns
zero_column_mask = torch.zeros([w_bitwidth, NUM_GROUP], dtype=torch.bool, device=device)
for i in range(w_bitwidth):
eq_zero = torch.all(torch.eq(wqb_signMagnitude[i], 0.), dim=-1)
zero_column_mask[i][eq_zero] = True
# prune_until is a pointer to specify which column to prune until
# E.g., for 8-bit weight -> column_idx = [0, 1, 2, 3, 4, 5, 6, 7]
# if require 4 zero columns, then we should prune from column 7 until column 4
prune_until = torch.full([NUM_GROUP], w_bitwidth-num_pruned_column, device=device)
# cHeck if there are zero columns before prune_until
for i in range(1, w_bitwidth):
mask = torch.logical_and(prune_until.gt(i), zero_column_mask[i])
prune_until[mask] += 1
# test_until is a pointer to specify which column to test until
test_until = torch.full([NUM_GROUP], 1, device=device)
for i in range(1, w_bitwidth):
mask = torch.logical_and(prune_until.gt(i), zero_column_mask[i])
test_until[mask] = i + 1
# prune columns [prune_until:], we should test columns [test_until:] to minimize MSE
# since the columns between test_until and prune_until can be adjusted arbitrarily as long as [prune_until:] are all zero
value_test = torch.zeros_like(wq_int, dtype=torch.float32, device=device)
value_new = torch.zeros_like(wq_int, dtype=torch.float32, device=device)
for test_idx in range(1, w_bitwidth):
mask = torch.eq(test_until, test_idx)
column_test = wqb_signMagnitude[test_idx:, mask, :]
int_test = binary_to_int(column_test, w_bitwidth=w_bitwidth-test_idx, device=device)
value_test[mask] = int_test
for prune_idx in range(w_bitwidth-num_pruned_column, w_bitwidth):
for test_idx in range(1, prune_idx):
mask_group = torch.logical_and(torch.eq(test_until, test_idx), torch.eq(prune_until, prune_idx))
mask_group = mask_group.unsqueeze(-1).expand(-1, group_size)
error = torch.full([NUM_GROUP, group_size], 1e7, device=device)
for n in range(2**(prune_idx-test_idx)):
tmp_value = n * 2.**(w_bitwidth-prune_idx)
new_error = (tmp_value - value_test) ** 2
mask_value = torch.logical_and(torch.lt(new_error, error), mask_group)
error[mask_value] = new_error[mask_value]
value_new[mask_value] = tmp_value
column_new = int_to_binary(value_new, w_bitwidth=w_bitwidth-test_idx, device=device)
wqb_signMagnitude[test_idx:, mask_group] = column_new[:, mask_group]
wq_int_new = signMagnitude_to_int(wqb_signMagnitude, w_bitwidth=w_bitwidth, device=device)
wq_int_new = wq_int_new.view(K, H, W, C).permute(0, 3, 1, 2)
return wq_int_new
def bitFlip_fc(wq_int, w_bitwidth: int=8, group_size: int=16, num_pruned_column: int=4, device='cpu'):
wq_int = wq_int.to(device)
K, C = wq_int.size() # output channel, input channel
if C < group_size:
group_size = C
NUM_GROUP = K*C//group_size
wq_int = wq_int.view(NUM_GROUP, group_size)
wqb_signMagnitude = int_to_signMagnitude(wq_int, w_bitwidth=w_bitwidth, device=device)
# check existing zero columns
zero_column_mask = torch.zeros([w_bitwidth, NUM_GROUP], dtype=torch.bool, device=device)
for i in range(w_bitwidth):
eq_zero = torch.all(torch.eq(wqb_signMagnitude[i], 0.), dim=-1)
zero_column_mask[i][eq_zero] = True
# prune_until is a pointer to specify which column to prune until
# E.g., for 8-bit weight -> column_idx = [0, 1, 2, 3, 4, 5, 6, 7]
# if require 4 zero columns, then we should prune from column 7 until column 4
prune_until = torch.full([NUM_GROUP], w_bitwidth-num_pruned_column, device=device)
# cHeck if there are zero columns before prune_until
for i in range(1, w_bitwidth):
mask = torch.logical_and(prune_until.gt(i), zero_column_mask[i])
prune_until[mask] += 1
# test_until is a pointer to specify which column to test until
test_until = torch.full([NUM_GROUP], 1, device=device)
for i in range(1, w_bitwidth):
mask = torch.logical_and(prune_until.gt(i), zero_column_mask[i])
test_until[mask] = i + 1
# prune columns [prune_until:], we should test columns [test_until:] to minimize MSE
# since the columns between test_until and prune_until can be adjusted arbitrarily as long as [prune_until:] are all zero
value_test = torch.zeros_like(wq_int, dtype=torch.float32, device=device)
value_new = torch.zeros_like(wq_int, dtype=torch.float32, device=device)
for test_idx in range(1, w_bitwidth):
mask = torch.eq(test_until, test_idx)
column_test = wqb_signMagnitude[test_idx:, mask, :]
int_test = binary_to_int(column_test, w_bitwidth=w_bitwidth-test_idx, device=device)
value_test[mask] = int_test
for prune_idx in range(w_bitwidth-num_pruned_column, w_bitwidth):
for test_idx in range(1, prune_idx):
mask_group = torch.logical_and(torch.eq(test_until, test_idx), torch.eq(prune_until, prune_idx))
mask_group = mask_group.unsqueeze(-1).expand(-1, group_size)
error = torch.full([NUM_GROUP, group_size], 1e7, device=device)
for n in range(2**(prune_idx-test_idx)):
tmp_value = n * 2.**(w_bitwidth-prune_idx)
new_error = (tmp_value - value_test) ** 2
mask_value = torch.logical_and(torch.lt(new_error, error), mask_group)
error[mask_value] = new_error[mask_value]
value_new[mask_value] = tmp_value
column_new = int_to_binary(value_new, w_bitwidth=w_bitwidth-test_idx, device=device)
wqb_signMagnitude[test_idx:, mask_group] = column_new[:, mask_group]
wq_int_new = signMagnitude_to_int(wqb_signMagnitude, w_bitwidth=w_bitwidth, device=device)
wq_int_new = wq_int_new.view(K, C)
return wq_int_new