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
|
/*******************************************************************************
* Copyright 2018 Intel Corporation
*
* 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.
*******************************************************************************/
/*
* Cell execution GRU with linear before reset
*/
#include "math_utils.hpp"
#include "mkldnn_thread.hpp"
#include "ref_rnn.hpp"
namespace mkldnn {
namespace impl {
namespace cpu {
using namespace mkldnn::impl::utils;
using namespace mkldnn::impl::math;
using namespace rnn_utils;
#define AOC array_offset_calculator
template <>
rnn_elemwise_sig(ref_rnn_fwd_f32_t::gru_lbr_elemwise) {
ws_gates_aoc_t ws_gates(rnn, ws_gates_);
bias_aoc_t bias(rnn, bias_);
ws_states_aoc_t states_t_l(rnn, states_t_l_);
ws_states_aoc_t states_tm1_l(rnn, states_tm1_l_);
ws_gates_aoc_t ws_gemm_state(rnn, ws_cell_);
AOC<float, 2> ws_Wh_b(ws_grid_, rnn.mb, rnn.dic);
parallel_nd(rnn.mb, [&](int i) {
PRAGMA_OMP_SIMD()
for (int j = 0; j < rnn.dic; j++) {
float Wh_b = ws_gemm_state(i, 2, j) + bias(3, j);
ws_gates(i, 0, j) = logistic_fwd(
ws_gates(i, 0, j) + ws_gemm_state(i, 0, j) + bias(0, j));
ws_gates(i, 1, j) = logistic_fwd(
ws_gates(i, 1, j) + ws_gemm_state(i, 1, j) + bias(1, j));
ws_gates(i, 2, j) = tanh_fwd(
ws_gates(i, 2, j) + ws_gates(i, 1, j) * Wh_b + bias(2, j));
states_t_l(i, j) = states_tm1_l(i, j) * ws_gates(i, 0, j)
+ (1.0f - ws_gates(i, 0, j)) * ws_gates(i, 2, j);
if (rnn.is_training)
ws_Wh_b(i, j) = Wh_b;
}
});
}
template <>
rnn_elemwise_sig(ref_rnn_fwd_u8s8_t::gru_lbr_elemwise) {
assert(!"GRU LBR int8 is not supported");
}
template <>
rnn_cell_execution_sig(ref_rnn_fwd_f32_t::cell_execution_gru_lbr) {
if (!rnn.merge_gemm_layer) {
(this->*gemm_layer_func)('N', 'N', rnn.n_gates * rnn.dic, rnn.mb,
rnn.slc, 1.0, w_layer_[0], rnn.weights_layer_ld,
states_t_lm1_, rnn.states_ws_ld, 0.0, ws_gates_,
rnn.gates_ws_ld);
}
(this->*gemm_iter_func)('N', 'N', rnn.n_gates * rnn.dic, rnn.mb, rnn.sic,
1.0, w_iter_[0], rnn.weights_iter_ld, states_tm1_l_,
rnn.states_ws_ld, 0.0, ws_cell_, rnn.gates_ws_ld);
(this->*elemwise_func)(rnn, ws_gates_, states_t_l_, c_states_t_l_,
states_tm1_l_, c_states_tm1_l_, diff_states_t_l_,
diff_states_t_lp1_, diff_states_tp1_l_, bias_[0], ws_grid_,
ws_cell_);
}
template <>
rnn_cell_execution_sig(ref_rnn_fwd_u8s8_t::cell_execution_gru_lbr) {
assert(!"GRU LBR int8 is not supported");
}
template <>
rnn_elemwise_sig(ref_rnn_bwd_f32_t::gru_lbr_elemwise) {
ws_gates_aoc_t ws_gates(rnn, ws_gates_);
ws_states_aoc_t states_tm1_l(rnn, states_tm1_l_);
ws_diff_states_aoc_t diff_states_t_l(rnn, diff_states_t_l_);
ws_diff_states_aoc_t diff_states_tp1_l(rnn, diff_states_tp1_l_);
ws_diff_states_aoc_t diff_states_t_lp1(rnn, diff_states_t_lp1_);
ws_gates_aoc_t ws_gates_r(rnn, ws_cell_);
AOC<float, 2> ws_Wh_b(ws_grid_, rnn.mb, rnn.dic);
// 1. calculate dG1 dG2 dG3
// dG0 = (dht - G2) * dht * (1 - G0) * G0
// dG1 = (W*h + b) * dG2 * (1 - G1) * G1
// dG2 = (1 - G0) * dht * (1 - G2*G2)
parallel_nd(rnn.mb, [&](int i) {
PRAGMA_OMP_SIMD()
for (int j = 0; j < rnn.dic; j++) {
float h = states_tm1_l(i, j);
float dHt = diff_states_tp1_l(0, i, j)
+ diff_states_t_lp1(rnn.n_states, i, j);
float dG0 = (h - ws_gates(i, 2, j)) * dHt
* x_m_square(ws_gates(i, 0, j));
float dG2 = (1.0f - ws_gates(i, 0, j))
* one_m_square(ws_gates(i, 2, j)) * dHt;
float dG1 = ws_Wh_b(i, j) * dG2 * x_m_square(ws_gates(i, 1, j));
diff_states_t_l(0, i, j) = dHt * ws_gates(i, 0, j);
ws_gates(i, 2, j) = dG2;
ws_gates_r(i, 2, j) = dG2 * ws_gates(i, 1, j);
ws_gates(i, 0, j) = ws_gates_r(i, 0, j) = dG0;
ws_gates(i, 1, j) = ws_gates_r(i, 1, j) = dG1;
}
});
}
template <>
rnn_cell_execution_sig(ref_rnn_bwd_f32_t::cell_execution_gru_lbr) {
ws_gates_aoc_t ws_gates_r(rnn, ws_cell_);
ws_diff_states_aoc_t diff_states_t_l(rnn, diff_states_t_l_);
(this->*elemwise_func)(rnn, ws_gates_, states_t_l_, c_states_t_l_,
states_tm1_l_, c_states_tm1_l_, diff_states_t_l_,
diff_states_t_lp1_, diff_states_tp1_l_, bias_[0], ws_grid_,
ws_cell_);
if (!rnn.merge_gemm_layer) {
// dx = dG * Wx^t
(this->*gemm_layer_func)('N', 'N', rnn.slc, rnn.mb,
rnn.n_gates * rnn.dic, 1.0, w_layer_[0],
rnn.weights_layer_ld, ws_gates_, rnn.gates_ws_ld, 0.0,
&diff_states_t_l(rnn.n_states, 0, 0), rnn.states_ws_ld);
// dWx += dG^t * x
gemm('N', 'T', rnn.n_gates * rnn.dic, rnn.slc, rnn.mb, 1.0, ws_gates_,
rnn.gates_ws_ld, states_t_lm1_, rnn.states_ws_ld, 1.0,
diff_w_layer_, rnn.diff_weights_layer_ld);
}
// dh += dGr * Wh^t
(this->*gemm_iter_func)('N', 'N', rnn.sic, rnn.mb, rnn.n_gates * rnn.dic,
1.0, w_iter_[0], rnn.weights_iter_ld, ws_cell_, rnn.gates_ws_ld,
1.0, diff_states_t_l_, rnn.states_ws_ld);
// dWh += dGr^t * h
gemm('N', 'T', rnn.n_gates * rnn.dic, rnn.sic, rnn.mb, 1.0, ws_cell_,
rnn.gates_ws_ld, states_tm1_l_, rnn.states_ws_ld, 1.0, diff_w_iter_,
rnn.diff_weights_layer_ld);
// db1-3 += e * dG
// db4 += e * (r * dG2)
gates_reduction(rnn, ws_gates_, diff_bias_);
parallel_nd(rnn.dic, [&](int j) {
for (int i = 0; i < rnn.mb; i++) {
diff_bias_[3 * rnn.dic + j] += ws_gates_r(i, 2, j);
}
});
}
#undef AOC
}
}
}
|