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

/*******************************************************************************
* 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
 */
#pragma warning(disable : 4503) /* name is too long */

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

#include "cpu/rnn/ref_rnn.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 <data_type_t src_type, data_type_t weights_type, data_type_t acc_type>
rnn_cell_execution_sig((ref_rnn_fwd_t<src_type, weights_type,
        acc_type>::cell_execution_gru_lbr)) {
    const auto src_layer_ld = rnn.src_layer_ld(cell_position);
    const auto src_iter_ld = rnn.src_iter_ld(cell_position);

    if (rnn.need_gemm_layer(cell_position)) {
        if (rnn.use_matmul) {
            CHECK(this->execute_matmul(ctx,
                    this->get_matmul_layer(cell_position), w_layer_[0],
                    src_layer_, scratch_gates_));
        } else {
            CHECK((this->*gemm_layer_func)('N', 'N', rnn.n_gates * rnn.dhc,
                    rnn.mb, rnn.slc, 1.0, w_layer_[0], rnn.weights_layer_ld,
                    src_layer_, src_layer_ld, 0.0, scratch_gates_,
                    rnn.scratch_gates_ld));
        }
    }
    if (rnn.use_matmul) {
        CHECK(this->execute_matmul(ctx, this->get_matmul_iter(cell_position),
                w_iter_[0], src_iter_, scratch_cell_));
    } else {
        CHECK((this->*gemm_iter_func)('N', 'N', rnn.n_gates * rnn.dhc, rnn.mb,
                rnn.sic, 1.0, w_iter_[0], rnn.weights_iter_ld, src_iter_,
                src_iter_ld, 0.0, scratch_cell_, rnn.scratch_gates_ld));
    }
    this->rnn_postgemm_->execute(rnn, cell_position, ws_gates_, scratch_gates_,
            augru_attention_, dst_layer_, dst_iter_c_, src_iter_, src_iter_c_,
            diff_src_layer_, diff_augru_attention_, diff_src_iter_,
            diff_src_iter_c_, diff_dst_layer_, diff_dst_iter_, nullptr, nullptr,
            bias_[0], ws_grid_, scratch_cell_, dst_iter_, nullptr, 0);

    return dnnl_success;
}

template rnn_cell_execution_sig(ref_rnn_fwd_f32_t::cell_execution_gru_lbr);
template rnn_cell_execution_sig(ref_rnn_fwd_bf16_t::cell_execution_gru_lbr);
template rnn_cell_execution_sig(ref_rnn_fwd_f16_t::cell_execution_gru_lbr);

template <>
rnn_cell_execution_sig(ref_rnn_fwd_u8s8_t::cell_execution_gru_lbr) {
    assert(!"GRU LBR int8 is not supported");
    return dnnl_unimplemented;
}

template <>
rnn_cell_execution_sig(ref_rnn_fwd_s8s8_t::cell_execution_gru_lbr) {
    assert(!"GRU LBR int8 is not supported");
    return dnnl_unimplemented;
}

template <typename T1, typename T2, typename T3, typename T4, typename T5,
        typename weights_data_t, typename src_data_t, typename acc_data_t,
        typename scratch_data_t>
dnnl_status_t common_bwd_cell_exec_template(T1 gemm_layer_f, T2 gemm_iter_f,
        T3 gemm_weights_layer_f, T4 gemm_weights_iter_f, T5 rnn_postgemm,
        const rnn_utils::rnn_conf_t &rnn, cell_position_t cell_position,
        src_data_t *dst_layer_, acc_data_t *diff_src_layer_,
        acc_data_t *diff_augru_attention_, acc_data_t *diff_src_iter_,
        weights_data_t **w_layer_, weights_data_t **w_iter_, void **bias_,
        const src_data_t *src_layer_, const src_data_t *augru_attention_,
        const src_data_t *src_iter_, acc_data_t *diff_dst_layer_,
        acc_data_t *diff_dst_iter_, acc_data_t *diff_w_layer_,
        acc_data_t *diff_w_iter_, acc_data_t *diff_bias_, src_data_t *ws_gates_,
        src_data_t *ws_grid_, scratch_data_t *scratch_gates_,
        scratch_data_t *scratch_cell_, src_data_t *dst_iter_) {
    const auto src_layer_ld = rnn.src_layer_ld(cell_position);
    const auto src_iter_ld = rnn.src_iter_ld(cell_position);

    const scratch_gates_aoc_t<scratch_data_t> scratch_gates_r(
            rnn, scratch_cell_);

    rnn_postgemm->execute(rnn, cell_position, ws_gates_, scratch_gates_,
            augru_attention_, dst_layer_, nullptr, src_iter_, nullptr, nullptr,
            diff_augru_attention_, diff_src_iter_, nullptr, diff_dst_layer_,
            diff_dst_iter_, nullptr, nullptr, bias_[0], ws_grid_, scratch_cell_,
            dst_iter_, nullptr, 0);

    // dWx +=  dG^t * x
    if (rnn.need_gemm_layer(cell_position))
        CHECK(gemm_weights_layer_f(
                scratch_gates_, src_layer_, src_layer_ld, diff_w_layer_));

    //  dx = dG * Wx^t
    if (!rnn.merge_gemm_layer)
        CHECK(gemm_layer_f(w_layer_[0], scratch_gates_, diff_src_layer_));

    // dh +=  dGr * Wh^t
    CHECK(gemm_iter_f(w_iter_[0], scratch_cell_, diff_src_iter_));

    // dWh += dGr^t * h
    CHECK(gemm_weights_iter_f(
            scratch_cell_, src_iter_, src_iter_ld, diff_w_iter_));

    // db1-3 += e * dG
    // db4 += e * (r * dG2)
    gates_reduction(rnn, cell_position, scratch_gates_, diff_bias_);

    parallel_nd(rnn.dhc, [&](dim_t j) {
        if (rnn.diff_weights_overwrite
                && (cell_position & rnn_utils::last_iter))
            diff_bias_[3 * rnn.dhc + j] = 0.0f;
        for (int i = 0; i < rnn.mb; i++) {
            diff_bias_[3 * rnn.dhc + j] += scratch_gates_r(i, 2, j);
        }
    });

    return dnnl_success;
}

#undef AOC

template <data_type_t src_type, data_type_t weights_type, data_type_t acc_type>
rnn_cell_execution_sig((ref_rnn_bwd_t<src_type, weights_type,
        acc_type>::cell_execution_gru_lbr)) {
    const auto gemm_layer
            = [&](const weights_t *A, const scratch_t *B, float *C) {
        return (this->*gemm_layer_func)('N', 'N', rnn.slc, rnn.mb,
                rnn.n_gates * rnn.dhc, 1.0f, A, rnn.weights_layer_ld, B,
                rnn.scratch_gates_ld, 0.0f, C, rnn.ws_diff_states_layer_ld);
    };
    const auto gemm_iter
            = [&](const weights_t *A, const scratch_t *B, float *C) {
        return (this->*gemm_iter_func)('N', 'N', rnn.sic, rnn.mb,
                rnn.n_gates * rnn.dhc, 1.0f, A, rnn.weights_iter_ld, B,
                rnn.ws_gates_ld, 1.0f, C, rnn.ws_diff_states_iter_ld);
    };
    const auto gemm_weights_layer
            = [&](const scratch_t *A, const src_layer_t *B, int ldb, float *C) {
        const float beta = rnn.diff_weights_beta(cell_position);
        return gemm('N', 'T', rnn.n_gates * rnn.dhc, rnn.slc, rnn.mb, 1.0f, A,
                rnn.scratch_gates_ld, B, ldb, beta, C,
                rnn.diff_weights_layer_ld);
    };
    const auto gemm_weights_iter
            = [&](const scratch_t *A, const src_iter_t *B, int ldb, float *C) {
        const float beta = rnn.diff_weights_beta(cell_position);
        return gemm('N', 'T', rnn.n_gates * rnn.dhc, rnn.sic, rnn.mb, 1.0f, A,
                rnn.ws_gates_ld, B, ldb, beta, C, rnn.diff_weights_iter_ld);
    };

    CHECK(common_bwd_cell_exec_template(gemm_layer, gemm_iter,
            gemm_weights_layer, gemm_weights_iter, this->rnn_postgemm_, rnn,
            cell_position, dst_layer_, diff_src_layer_, diff_augru_attention_,
            diff_src_iter_, w_layer_, w_iter_, bias_, src_layer_,
            augru_attention_, src_iter_, diff_dst_layer_, diff_dst_iter_,
            diff_w_layer_, diff_w_iter_, diff_bias_, ws_gates_, ws_grid_,
            scratch_gates_, scratch_cell_, dst_iter_));
    return dnnl_success;
}

template rnn_cell_execution_sig(ref_rnn_bwd_f32_t::cell_execution_gru_lbr);
template rnn_cell_execution_sig(ref_rnn_bwd_bf16_t::cell_execution_gru_lbr);
template rnn_cell_execution_sig(ref_rnn_bwd_f16_t::cell_execution_gru_lbr);

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