ncnn/src/layer/convolutiondepthwise.cpp

// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2017 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// 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.

#include "convolutiondepthwise.h"

#include "layer_type.h"

#include "fused_activation.h"

namespace ncnn {

ConvolutionDepthWise::ConvolutionDepthWise()
{
    one_blob_only = true;
    support_inplace = false;
}

int ConvolutionDepthWise::load_param(const ParamDict& pd)
{
    num_output = pd.get(0, 0);
    kernel_w = pd.get(1, 0);
    kernel_h = pd.get(11, kernel_w);
    dilation_w = pd.get(2, 1);
    dilation_h = pd.get(12, dilation_w);
    stride_w = pd.get(3, 1);
    stride_h = pd.get(13, stride_w);
    pad_left = pd.get(4, 0);
    pad_right = pd.get(15, pad_left);
    pad_top = pd.get(14, pad_left);
    pad_bottom = pd.get(16, pad_top);
    pad_value = pd.get(18, 0.f);
    bias_term = pd.get(5, 0);
    weight_data_size = pd.get(6, 0);
    group = pd.get(7, 1);
    int8_scale_term = pd.get(8, 0);
    activation_type = pd.get(9, 0);
    activation_params = pd.get(10, Mat());

    dynamic_weight = pd.get(19, 0);

    if (dynamic_weight)
    {
        one_blob_only = false;
    }

    if (num_output % group != 0)
    {
        // reject invalid group
        return -100;
    }

    if (int8_scale_term)
    {
#if NCNN_INT8
        support_int8_storage = true;
#else
        NCNN_LOGE("please build ncnn with NCNN_INT8 enabled for int8 inference");
        return -1;
#endif
    }

    return 0;
}

int ConvolutionDepthWise::load_model(const ModelBin& mb)
{
    if (dynamic_weight)
        return 0;

    weight_data = mb.load(weight_data_size, 0);
    if (weight_data.empty())
        return -100;

    if (bias_term)
    {
        bias_data = mb.load(num_output, 1);
        if (bias_data.empty())
            return -100;
    }

#if NCNN_INT8
    if (int8_scale_term == 1 || int8_scale_term == 101)
    {
        weight_data_int8_scales = mb.load(group, 1);
        bottom_blob_int8_scales = mb.load(1, 1);

        float bottom_blob_int8_scale = bottom_blob_int8_scales[0];
        bottom_blob_int8_scales = Mat(group);
        bottom_blob_int8_scales.fill(bottom_blob_int8_scale);
    }
    else if (int8_scale_term == 2 || int8_scale_term == 102)
    {
        weight_data_int8_scales = mb.load(1, 1);
        bottom_blob_int8_scales = mb.load(1, 1);

        // extend group if only one provided
        float weight_data_int8_scale = weight_data_int8_scales[0];
        weight_data_int8_scales = Mat(group);
        weight_data_int8_scales.fill(weight_data_int8_scale);

        float bottom_blob_int8_scale = bottom_blob_int8_scales[0];
        bottom_blob_int8_scales = Mat(group);
        bottom_blob_int8_scales.fill(bottom_blob_int8_scale);
    }

    if (int8_scale_term > 100)
    {
        top_blob_int8_scales = mb.load(1, 1);

        float top_blob_int8_scale = top_blob_int8_scales[0];
        top_blob_int8_scales = Mat(group);
        top_blob_int8_scales.fill(top_blob_int8_scale);
    }
#endif // NCNN_INT8

#if NCNN_INT8
    // runtime quantize the weight data
    if (weight_data.elemsize == (size_t)4u && int8_scale_term)
    {
        Mat int8_weight_data(weight_data_size, (size_t)1u);
        if (int8_weight_data.empty())
            return -100;

        const int weight_data_size_g = weight_data_size / group;

        for (int g = 0; g < group; g++)
        {
            Option opt_q;
            opt_q.num_threads = 1;
            opt_q.blob_allocator = int8_weight_data.allocator;
            opt_q.use_packing_layout = false;

            const Mat weight_data_g = weight_data.range(weight_data_size_g * g, weight_data_size_g);
            Mat int8_weight_data_g = int8_weight_data.range(weight_data_size_g * g, weight_data_size_g);
            const Mat weight_data_int8_scales_g = weight_data_int8_scales.range(g, 1);
            quantize_to_int8(weight_data_g, int8_weight_data_g, weight_data_int8_scales_g, opt_q);
        }

        weight_data = int8_weight_data;
    }
#endif // NCNN_INT8

    return 0;
}

static int convolutiondepthwise(const Mat& bottom_blob, Mat& top_blob, const Mat& weight_data, const Mat& bias_data, int kernel_w, int kernel_h, int stride_w, int stride_h, int dilation_w, int dilation_h, int group, int activation_type, const Mat& activation_params, const Option& opt)
{
    const int w = bottom_blob.w;
    const int inch = bottom_blob.c;

    const int outw = top_blob.w;
    const int outh = top_blob.h;
    const int outch = top_blob.c;

    const int bias_term = bias_data.empty() ? 0 : 1;

    const int maxk = kernel_w * kernel_h;

    // kernel offsets
    std::vector<int> _space_ofs(maxk);
    int* space_ofs = &_space_ofs[0];
    {
        int p1 = 0;
        int p2 = 0;
        int gap = w * dilation_h - kernel_w * dilation_w;
        for (int i = 0; i < kernel_h; i++)
        {
            for (int j = 0; j < kernel_w; j++)
            {
                space_ofs[p1] = p2;
                p1++;
                p2 += dilation_w;
            }
            p2 += gap;
        }
    }

    // depth-wise
    if (inch == group && group == outch)
    {
        #pragma omp parallel for num_threads(opt.num_threads)
        for (int g = 0; g < group; g++)
        {
            float* outptr = top_blob.channel(g);
            const float* kptr = (const float*)weight_data + maxk * g;
            const Mat m = bottom_blob.channel(g);

            for (int i = 0; i < outh; i++)
            {
                for (int j = 0; j < outw; j++)
                {
                    float sum = 0.f;

                    if (bias_term)
                        sum = bias_data[g];

                    const float* sptr = m.row(i * stride_h) + j * stride_w;

                    for (int k = 0; k < maxk; k++)
                    {
                        float val = sptr[space_ofs[k]];
                        float w = kptr[k];
                        sum += val * w;
                    }

                    outptr[j] = activation_ss(sum, activation_type, activation_params);
                }

                outptr += outw;
            }
        }
    }
    else
    {
        // group convolution
        const int inch_g = inch / group;
        const int outch_g = outch / group;

#ifdef _WIN32
        #pragma omp parallel for num_threads(opt.num_threads)
#else
        #pragma omp parallel for collapse(2) num_threads(opt.num_threads)
#endif
        for (int g = 0; g < group; g++)
        {
            for (int p = 0; p < outch_g; p++)
            {
                float* outptr = top_blob.channel(g * outch_g + p);
                const float* weight_data_ptr = (const float*)weight_data + maxk * inch_g * outch_g * g;

                // shadowed variable for less openmp task args
                const int outw = top_blob.w;
                const int outh = top_blob.h;

                for (int i = 0; i < outh; i++)
                {
                    for (int j = 0; j < outw; j++)
                    {
                        float sum = 0.f;

                        if (bias_term)
                            sum = bias_data[outch_g * g + p];

                        const float* kptr = weight_data_ptr + maxk * inch_g * p;

                        for (int q = 0; q < inch_g; q++)
                        {
                            const Mat m = bottom_blob.channel(inch_g * g + q);
                            const float* sptr = m.row(i * stride_h) + j * stride_w;

                            for (int k = 0; k < maxk; k++)
                            {
                                float val = sptr[space_ofs[k]];
                                float w = kptr[k];
                                sum += val * w;
                            }

                            kptr += maxk;
                        }

                        outptr[j] = activation_ss(sum, activation_type, activation_params);
                    }

                    outptr += outw;
                }
            }
        }
    }

    return 0;
}

int ConvolutionDepthWise::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
{
    // convolv with NxN kernel
    // value = value + bias

#if NCNN_INT8
    if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u)
    {
        return forward_int8(bottom_blob, top_blob, opt);
    }
#endif

    //     NCNN_LOGE("ConvolutionDepthWise input %d x %d  pad = %d %d  ksize=%d %d  stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h);

    Mat bottom_blob_bordered;
    make_padding(bottom_blob, bottom_blob_bordered, opt);
    if (bottom_blob_bordered.empty())
        return -100;

    const int w = bottom_blob_bordered.w;
    const int h = bottom_blob_bordered.h;
    const size_t elemsize = bottom_blob_bordered.elemsize;

    const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1;
    const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1;

    const int outw = (w - kernel_extent_w) / stride_w + 1;
    const int outh = (h - kernel_extent_h) / stride_h + 1;

    top_blob.create(outw, outh, num_output, elemsize, opt.blob_allocator);
    if (top_blob.empty())
        return -100;

    int ret = convolutiondepthwise(bottom_blob_bordered, top_blob, weight_data, bias_data, kernel_w, kernel_h, stride_w, stride_h, dilation_w, dilation_h, group, activation_type, activation_params, opt);
    if (ret != 0)
        return ret;

    return 0;
}

int ConvolutionDepthWise::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const
{
    const Mat& bottom_blob = bottom_blobs[0];
    const Mat& _weight_data = bottom_blobs[1];
    Mat& top_blob = top_blobs[0];

    const int _kernel_w = _weight_data.w;
    const int _kernel_h = _weight_data.h;
    const int _num_output = _weight_data.c;

    Mat weight_data_flattened;
    flatten(_weight_data, weight_data_flattened, opt);
    if (weight_data_flattened.empty())
        return -100;

    Mat bias_data_flattened;
    if (bias_term)
    {
        const Mat& _bias_data = bottom_blobs[2];
        flatten(_bias_data, bias_data_flattened, opt);
        if (bias_data_flattened.empty())
            return -100;
    }

    Mat bottom_blob_bordered;
    make_padding(bottom_blob, bottom_blob_bordered, _kernel_w, _kernel_h, opt);
    if (bottom_blob_bordered.empty())
        return -100;

    const int w = bottom_blob_bordered.w;
    const int h = bottom_blob_bordered.h;
    const size_t elemsize = bottom_blob_bordered.elemsize;

    const int kernel_extent_w = dilation_w * (_kernel_w - 1) + 1;
    const int kernel_extent_h = dilation_h * (_kernel_h - 1) + 1;

    const int outw = (w - kernel_extent_w) / stride_w + 1;
    const int outh = (h - kernel_extent_h) / stride_h + 1;

    top_blob.create(outw, outh, _num_output, elemsize, opt.blob_allocator);
    if (top_blob.empty())
        return -100;

    int ret = convolutiondepthwise(bottom_blob_bordered, top_blob, weight_data_flattened, bias_data_flattened, _kernel_w, _kernel_h, stride_w, stride_h, dilation_w, dilation_h, group, activation_type, activation_params, opt);
    if (ret != 0)
        return ret;

    return 0;
}

void ConvolutionDepthWise::make_padding(const Mat& bottom_blob, Mat& bottom_blob_bordered, const Option& opt) const
{
    make_padding(bottom_blob, bottom_blob_bordered, kernel_w, kernel_h, opt);
}

void ConvolutionDepthWise::make_padding(const Mat& bottom_blob, Mat& bottom_blob_bordered, int _kernel_w, int _kernel_h, const Option& opt) const
{
    int w = bottom_blob.w;
    int h = bottom_blob.h;

    const int kernel_extent_w = dilation_w * (_kernel_w - 1) + 1;
    const int kernel_extent_h = dilation_h * (_kernel_h - 1) + 1;

    bottom_blob_bordered = bottom_blob;
    if (pad_left > 0 || pad_right > 0 || pad_top > 0 || pad_bottom > 0)
    {
        Option opt_b = opt;
        opt_b.blob_allocator = opt.workspace_allocator;
        copy_make_border(bottom_blob, bottom_blob_bordered, pad_top, pad_bottom, pad_left, pad_right, BORDER_CONSTANT, pad_value, opt_b);
    }
    else if (pad_left == -233 && pad_right == -233 && pad_top == -233 && pad_bottom == -233)
    {
        // tensorflow padding=SAME or onnx padding=SAME_UPPER
        int wpad = kernel_extent_w + (w - 1) / stride_w * stride_w - w;
        int hpad = kernel_extent_h + (h - 1) / stride_h * stride_h - h;
        if (wpad > 0 || hpad > 0)
        {
            Option opt_b = opt;
            opt_b.blob_allocator = opt.workspace_allocator;
            copy_make_border(bottom_blob, bottom_blob_bordered, hpad / 2, hpad - hpad / 2, wpad / 2, wpad - wpad / 2, BORDER_CONSTANT, pad_value, opt_b);
        }
    }
    else if (pad_left == -234 && pad_right == -234 && pad_top == -234 && pad_bottom == -234)
    {
        // onnx padding=SAME_LOWER
        int wpad = kernel_extent_w + (w - 1) / stride_w * stride_w - w;
        int hpad = kernel_extent_h + (h - 1) / stride_h * stride_h - h;
        if (wpad > 0 || hpad > 0)
        {
            Option opt_b = opt;
            opt_b.blob_allocator = opt.workspace_allocator;
            copy_make_border(bottom_blob, bottom_blob_bordered, hpad - hpad / 2, hpad / 2, wpad - wpad / 2, wpad / 2, BORDER_CONSTANT, pad_value, opt_b);
        }
    }
}

#if NCNN_INT8
static inline signed char float2int8(float v)
{
    int int32 = static_cast<int>(round(v));
    if (int32 > 127) return 127;
    if (int32 < -127) return -127;
    return (signed char)int32;
}

int ConvolutionDepthWise::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
{
    // convolv with NxN kernel
    // value = value + bias

    int w = bottom_blob.w;
    int h = bottom_blob.h;
    int channels = bottom_blob.c;
    size_t elemsize = bottom_blob.elemsize;

    if (channels % group != 0 || num_output % group != 0)
    {
        // reject invalid group
        return -100;
    }

    //     NCNN_LOGE("ConvolutionDepthWise input %d x %d  pad = %d %d  ksize=%d %d  stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h);

    const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1;
    const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1;

    Mat bottom_blob_int8 = bottom_blob;
    if (elemsize != 1)
    {
        const int channels_g = channels / group;

        Mat scales(channels);
        {
            float* ps = scales;
            for (int g = 0; g < group; g++)
            {
                float scale = bottom_blob_int8_scales[g];
                for (int q = 0; q < channels_g; q++)
                {
                    *ps++ = scale;
                }
            }
        }

        Option opt_q = opt;
        opt_q.blob_allocator = opt.workspace_allocator;
        quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q);
    }

    Mat bottom_blob_bordered;
    make_padding(bottom_blob_int8, bottom_blob_bordered, opt);
    if (bottom_blob_bordered.empty())
        return -100;

    w = bottom_blob_bordered.w;
    h = bottom_blob_bordered.h;

    int outw = (w - kernel_extent_w) / stride_w + 1;
    int outh = (h - kernel_extent_h) / stride_h + 1;

    const int maxk = kernel_w * kernel_h;

    // kernel offsets
    std::vector<int> _space_ofs(maxk);
    int* space_ofs = &_space_ofs[0];
    {
        int p1 = 0;
        int p2 = 0;
        int gap = w * dilation_h - kernel_w * dilation_w;
        for (int i = 0; i < kernel_h; i++)
        {
            for (int j = 0; j < kernel_w; j++)
            {
                space_ofs[p1] = p2;
                p1++;
                p2 += dilation_w;
            }
            p2 += gap;
        }
    }

    // int8
    bool use_int8_requantize = int8_scale_term > 100;
    size_t out_elemsize = use_int8_requantize ? 1u : 4u;

    top_blob.create(outw, outh, num_output, out_elemsize, opt.blob_allocator);
    if (top_blob.empty())
        return -100;

    // depth-wise
    if (channels == group && group == num_output)
    {
        #pragma omp parallel for num_threads(opt.num_threads)
        for (int g = 0; g < group; g++)
        {
            signed char* outptr = top_blob.channel(g);
            const signed char* kptr = (const signed char*)weight_data + maxk * g;
            const Mat m = bottom_blob_bordered.channel(g);

            for (int i = 0; i < outh; i++)
            {
                for (int j = 0; j < outw; j++)
                {
                    int sum = 0;

                    const signed char* sptr = m.row<signed char>(i * stride_h) + j * stride_w;

                    for (int k = 0; k < maxk; k++)
                    {
                        signed char val = sptr[space_ofs[k]];
                        signed char w = kptr[k];
                        sum += val * w;
                    }

                    float scale_in;
                    if (weight_data_int8_scales[g] == 0)
                        scale_in = 0;
                    else
                        scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]);

                    float sumfp32 = sum * scale_in;

                    if (bias_term)
                        sumfp32 += bias_data[g];

                    sumfp32 = activation_ss(sumfp32, activation_type, activation_params);

                    if (use_int8_requantize)
                    {
                        // requantize
                        float scale_out = top_blob_int8_scales[g];
                        signed char sums8 = float2int8(sumfp32 * scale_out);
                        outptr[0] = sums8;
                        outptr += 1;
                    }
                    else
                    {
                        // dequantize
                        ((float*)outptr)[0] = sumfp32;
                        outptr += 4;
                    }
                }
            }
        }
    }
    else
    {
        // group convolution
        const int channels_g = channels / group;
        const int num_output_g = num_output / group;

#ifdef _WIN32
        #pragma omp parallel for num_threads(opt.num_threads)
#else // _WIN32
        #pragma omp parallel for collapse(2) num_threads(opt.num_threads)
#endif // _WIN32
        for (int g = 0; g < group; g++)
        {
            for (int p = 0; p < num_output_g; p++)
            {
                signed char* outptr = top_blob.channel(g * num_output_g + p);
                const signed char* weight_data_ptr = (const signed char*)weight_data + maxk * channels_g * num_output_g * g;

                for (int i = 0; i < outh; i++)
                {
                    for (int j = 0; j < outw; j++)
                    {
                        int sum = 0;

                        const signed char* kptr = weight_data_ptr + maxk * channels_g * p;

                        // channels_g
                        for (int q = 0; q < channels_g; q++)
                        {
                            const Mat m = bottom_blob_bordered.channel(channels_g * g + q);
                            const signed char* sptr = m.row<signed char>(i * stride_h) + j * stride_w;

                            for (int k = 0; k < maxk; k++)
                            {
                                signed char val = sptr[space_ofs[k]];
                                signed char w = kptr[k];
                                sum += val * w;
                            }

                            kptr += maxk;
                        }

                        float scale_in;
                        if (weight_data_int8_scales[g] == 0)
                            scale_in = 0;
                        else
                            scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]);

                        float sumfp32 = sum * scale_in;

                        if (bias_term)
                            sumfp32 += bias_data[g * num_output_g + p];

                        sumfp32 = activation_ss(sumfp32, activation_type, activation_params);

                        if (use_int8_requantize)
                        {
                            // requantize
                            float scale_out = top_blob_int8_scales[g];
                            signed char sums8 = float2int8(sumfp32 * scale_out);
                            outptr[0] = sums8;
                            outptr += 1;
                        }
                        else
                        {
                            // dequantize
                            ((float*)outptr)[0] = sumfp32;
                            outptr += 4;
                        }
                    }
                }
            }
        }
    }

    return 0;
}
#endif // NCNN_INT8

} // namespace ncnn