onednn-src 0.1.13

Source of oneAPI Deep Neural Network Library (oneDNN)
Documentation 
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

/*******************************************************************************
* 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 LSTM
 */

#include "common/dnnl_thread.hpp"
#include "common/math_utils.hpp"

#include "cpu/rnn/postgemm_dispatcher.hpp"

namespace dnnl {
namespace impl {
namespace cpu {

using namespace dnnl::impl::utils;
using namespace dnnl::impl::math;
using namespace rnn_utils;
#define AOC array_offset_calculator

template <typename T1, typename T2, typename T3, typename src_data_t,
        typename scratch_data_t>
void gru_lbr_fwd_postgemm_template(T1 func1, T2 func2, T3 to_src,
        const float *scales, const rnn_utils::rnn_conf_t &rnn,
        rnn_utils::cell_position_t cell_position, src_data_t *ws_gates_,
        scratch_data_t *scratch_gates_, const src_data_t *augru_attention_,
        src_data_t *dst_layer_, src_data_t *dst_iter_,
        const src_data_t *src_iter_, const void *bias_, src_data_t *ws_grid_,
        scratch_data_t *scratch_cell_, int block_step) {

    const auto src_iter_ld = rnn.src_iter_ld(cell_position);
    const auto dst_layer_ld = rnn.dst_layer_ld(cell_position);
    const auto dst_iter_ld = rnn.dst_iter_ld(cell_position);

    const augru_attention_aoc_t<const src_data_t> augru_attention(
            rnn, augru_attention_);
    const ws_states_layer_aoc_t<src_data_t> dst_layer(
            rnn, dst_layer_, dst_layer_ld);
    const ws_states_iter_aoc_t<src_data_t> dst_iter(
            rnn, dst_iter_, dst_iter_ld);
    const ws_states_iter_aoc_t<const src_data_t> src_iter(
            rnn, src_iter_, src_iter_ld);
    const ws_gates_aoc_t<src_data_t> ws_gates(rnn, ws_gates_);
    const scratch_gates_aoc_t<scratch_data_t> scratch_gates(
            rnn, scratch_gates_);
    const auto bias_aoc = rnn_utils::make_raw_aoc(
            bias_, types::data_type_size(rnn.bias_dt), rnn.n_bias, rnn.dhc);

    const auto bias = [&](int gate_id, int dhc_id) {
        return to_float(bias_aoc(gate_id, dhc_id), rnn.bias_dt);
    };
    const scratch_gates_aoc_t<scratch_data_t> scratch_cell(rnn, scratch_cell_);
    const AOC<src_data_t, 2> ws_Wh_b(ws_grid_, rnn.mb, rnn.dhc);

    const auto get_scales = [](const float *scales, int idx) {
        return scales ? scales + idx : nullptr;
    };
    const float *scales_G1 = get_scales(scales, 1);
    const float *scales_G2 = get_scales(scales, 2);

    const auto postgemm = [&](dim_t i) {
        PRAGMA_OMP_SIMD()
        for (int j = 0; j < rnn.dhc; j++) {
            const float Wh_b = scratch_cell(i, 2, j) + bias(3, j);
            auto G0 = func1(scales, // default func1 is sigmoid
                    scratch_gates(i, 0, j) + scratch_cell(i, 0, j)
                            + bias(0, j));
            const auto G1 = func1(scales_G1, // default func1 is sigmoid
                    scratch_gates(i, 1, j) + scratch_cell(i, 1, j)
                            + bias(1, j));
            const auto G2 = func2(scales_G2, // default func2 is tanh
                    scratch_gates(i, 2, j) + G1 * Wh_b + bias(2, j));
            if (rnn.is_training) {
                ws_gates(i, 0, j) = to_src(G0);
                ws_gates(i, 1, j) = to_src(G1);
                ws_gates(i, 2, j) = to_src(G2);
                ws_Wh_b(i, j) = to_src(Wh_b);
            }
            if (rnn.is_augru) {
                const auto a = to_src(augru_attention(i));
                G0 = (1.0f - a) * G0;
            }
            const auto tmp = to_src(src_iter(i, j) * G0 + (1.0f - G0) * G2);
            if (dst_layer_ != nullptr) dst_layer(i, j) = tmp;
            if (dst_iter_ != nullptr) dst_iter(i, j) = tmp;
        }
    };

    const auto postgemm_brgemm = [&](dim_t i) {
        const int n_elem = block_step;
        PRAGMA_OMP_SIMD()
        for (int j = 0; j < n_elem; j++) {
            const float Wh_b = scratch_cell(i, 0, j) + bias(3, j);
            auto G0 = func1(scales, // default func1 is sigmoid
                    scratch_gates(i, 0, j) + bias(0, j));
            const auto G1 = func1(scales_G1, // default func1 is sigmoid
                    scratch_gates(i, 1, j) + bias(1, j));
            const auto G2 = func2(scales_G2, // default func2 is tanh
                    scratch_gates(i, 2, j) + G1 * Wh_b + bias(2, j));
            if (rnn.is_training) {
                ws_gates(i, 0, j) = to_src(G0);
                ws_gates(i, 1, j) = to_src(G1);
                ws_gates(i, 2, j) = to_src(G2);
                ws_Wh_b(i, j) = to_src(Wh_b);
            }
            if (rnn.is_augru) {
                const auto a = to_src(augru_attention(i));
                G0 = (1.0f - a) * G0;
            }
            const auto tmp = to_src(src_iter(i, j) * G0 + (1.0f - G0) * G2);
            if (dst_layer_ != nullptr) dst_layer(i, j) = tmp;
            if (dst_iter_ != nullptr) dst_iter(i, j) = tmp;
        }
    };

    if (rnn.is_brgemm && !rnn.unfused_post_gemm) {
        for (dim_t i = 0; i < rnn.m_block; i++)
            postgemm_brgemm(i);
    } else {
        parallel_nd(rnn.mb, [&](dim_t i) { postgemm(i); });
    }
}

template <data_type_t src_type, data_type_t scratch_type, data_type_t acc_type>
rnn_postgemm_sig((rnn_postgemm_fwd_t<src_type, scratch_type,
        acc_type>::gru_lbr_postgemm)) {
    const float *scales = this->pd_->attr()->rnn_tparams_.scales_;

    const auto linear_f
            = [](const float *scale, float a) { return *scale * a; };
    const auto logistic_f = [](const float *scale, float a) {
        return logistic_fwd<float>(a);
    };
    const auto tanh_f
            = [](const float *scale, float a) { return tanh_fwd<float>(a); };
    const auto to_src = [](float a) { return gates_t(a); };

    if (!this->pd_->attr()->rnn_tparams_.test_mode_)
        gru_lbr_fwd_postgemm_template(logistic_f, tanh_f, to_src, scales, rnn,
                cell_position, ws_gates_, scratch_gates_, augru_attention_,
                dst_layer_, dst_iter_, src_iter_, bias_, ws_grid_,
                scratch_cell_, block_step);
    else
        gru_lbr_fwd_postgemm_template(linear_f, linear_f, to_src, scales, rnn,
                cell_position, ws_gates_, scratch_gates_, augru_attention_,
                dst_layer_, dst_iter_, src_iter_, bias_, ws_grid_,
                scratch_cell_, block_step);
}

template <>
rnn_postgemm_sig(rnn_postgemm_fwd_u8_t::gru_lbr_postgemm) {
    assert(!"GRU LBR int8 is not supported");
}

template <>
rnn_postgemm_sig(rnn_postgemm_fwd_s8_t::gru_lbr_postgemm) {
    assert(!"GRU LBR signed int8 is not supported");
}

template rnn_postgemm_sig(rnn_postgemm_fwd_f32_t::gru_lbr_postgemm);
template rnn_postgemm_sig(rnn_postgemm_fwd_bf16_t::gru_lbr_postgemm);
template rnn_postgemm_sig(rnn_postgemm_fwd_f16_t::gru_lbr_postgemm);

template <typename T1, typename src_data_t, typename acc_data_t,
        typename scratch_data_t>
void gru_lbr_bwd_postgemm_template(T1 to_src, const rnn_utils::rnn_conf_t &rnn,
        cell_position_t cell_position, src_data_t *ws_gates_,
        scratch_data_t *scratch_gates_, const src_data_t *augru_attention_,
        const src_data_t *src_iter_, acc_data_t *diff_src_iter_,
        acc_data_t *diff_dst_iter_, acc_data_t *diff_augru_attention_,
        acc_data_t *diff_dst_layer_, scratch_data_t *scratch_cell_,
        src_data_t *ws_grid_) {
    const auto src_iter_ld = rnn.src_iter_ld(cell_position);

    const augru_attention_aoc_t<const src_data_t> augru_attention(
            rnn, augru_attention_);
    const augru_attention_aoc_t<acc_data_t> diff_augru_attention(
            rnn, diff_augru_attention_);

    const ws_states_iter_aoc_t<const src_data_t> src_iter(
            rnn, src_iter_, src_iter_ld);
    const ws_gates_aoc_t<src_data_t> ws_gates(rnn, ws_gates_);
    const scratch_gates_aoc_t<scratch_data_t> scratch_gates(
            rnn, scratch_gates_);
    const ws_diff_states_iter_aoc_t<acc_data_t> diff_src_iter(
            rnn, diff_src_iter_);
    const ws_diff_states_iter_aoc_t<acc_data_t> diff_dst_iter(
            rnn, diff_dst_iter_);
    const ws_diff_states_layer_aoc_t<acc_data_t> diff_dst_layer(
            rnn, diff_dst_layer_);
    const scratch_gates_aoc_t<scratch_data_t> scratch_gates_r(
            rnn, scratch_cell_);
    const AOC<src_data_t, 2> ws_Wh_b(ws_grid_, rnn.mb, rnn.dhc);

    // 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, [&](dim_t i) {
        acc_data_t diff_attention = 0.0f;
        PRAGMA_OMP_SIMD(reduction(+ : diff_attention))
        for (int j = 0; j < rnn.dhc; j++) {
            const float h = src_iter(i, j);
            const float dHt = diff_dst_iter(i, j) + diff_dst_layer(i, j);
            float dG0 = (h - ws_gates(i, 2, j)) * dHt
                    * x_m_square(ws_gates(i, 0, j));
            const float dG2 = (1.0f - ws_gates(i, 0, j))
                    * one_m_square(ws_gates(i, 2, j)) * dHt;
            const float dG1
                    = ws_Wh_b(i, j) * dG2 * x_m_square(ws_gates(i, 1, j));

            if (rnn.is_augru) {
                diff_attention -= dG0 * ws_gates(i, 0, j);
                dG0 *= 1.0f - augru_attention(i);
            }

            diff_src_iter(i, j) = dHt * ws_gates(i, 0, j);
            scratch_gates(i, 2, j) = to_src(dG2);
            scratch_gates_r(i, 2, j) = to_src(dG2 * ws_gates(i, 1, j));
            scratch_gates(i, 0, j) = scratch_gates_r(i, 0, j) = to_src(dG0);
            scratch_gates(i, 1, j) = scratch_gates_r(i, 1, j) = to_src(dG1);
        }
        if (rnn.is_augru) diff_augru_attention(i) = diff_attention;
    });
}

template <data_type_t src_type, data_type_t scratch_type, data_type_t acc_type>
rnn_postgemm_sig((rnn_postgemm_bwd_t<src_type, scratch_type,
        acc_type>::gru_lbr_postgemm)) {
    auto to_src = [&](float a) { return scratch_t(a); };
    gru_lbr_bwd_postgemm_template(to_src, rnn, cell_position, ws_gates_,
            scratch_gates_, augru_attention_, src_iter_, diff_src_iter_,
            diff_dst_iter_, diff_augru_attention_, diff_dst_layer_,
            scratch_cell_, ws_grid_);
}

template rnn_postgemm_sig(rnn_postgemm_bwd_f32_t::gru_lbr_postgemm);
template rnn_postgemm_sig(rnn_postgemm_bwd_bf16_t::gru_lbr_postgemm);
template rnn_postgemm_sig(rnn_postgemm_bwd_f16_t::gru_lbr_postgemm);

} // namespace cpu
} // namespace impl
} // namespace dnnl