/******************************************************************************* * Copyright (c) 2015-2018 Skymind, Inc. * * This program and the accompanying materials are made available under the * terms of the Apache License, Version 2.0 which is available at * https://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. * * SPDX-License-Identifier: Apache-2.0 ******************************************************************************/ // // @author saudet // @author raver119@gmail.com // #include #include #include #include #include "mkldnnUtils.h" #include using namespace dnnl; namespace nd4j { namespace ops { namespace platforms { PLATFORM_IMPL(maxpool2d_bp) { auto input = INPUT_VARIABLE( 0); // [bS, iH, iW, iC] (NHWC) or [bS, iC, iH, iW] (NCHW) auto gradO = INPUT_VARIABLE( 1); // [bS, oH, oW, oC] (NHWC) or [bS, oC, oH, oW] (NCHW), epsilon_next auto gradI = OUTPUT_VARIABLE( 0); // [bS, iH, iW, iC] (NHWC) or [bS, iC, iH, iW] (NCHW), epsilon int kH = INT_ARG(0); // filter(kernel) height int kW = INT_ARG(1); // filter(kernel) width int sH = INT_ARG(2); // strides height int sW = INT_ARG(3); // strides width int pH = INT_ARG(4); // paddings height int pW = INT_ARG(5); // paddings width int dH = INT_ARG(6); // dilations height int dW = INT_ARG(7); // dilations width int isSameMode = INT_ARG(8); // 0-VALID, 1-SAME int extraParam0 = INT_ARG(9); int isNCHW = block.getIArguments()->size() > 10 ? !INT_ARG(10) : 1; // INT_ARG(10): 0-NCHW, 1-NHWC REQUIRE_TRUE(input->rankOf() == 4, 0, "AVGPOOL2D_BP op: input should have rank of 4, but got %i instead", input->rankOf()); REQUIRE_TRUE(dH != 0 && dW != 0, 0, "AVGPOOL2D_BP op: dilation must not be zero, but got instead {%i, %i}", dH, dW); int bS, iC, iH, iW, oC, oH, oW; // batch size, input channels, input height/width, output channels, output height/width; int indIOioC, indIiH, indWoC, indWiC, indWkH, indOoH; // corresponding indexes ConvolutionUtils::getSizesAndIndexesConv2d(isNCHW, *input, *gradO, bS, iC, iH, iW, oC, oH, oW, indIOioC, indIiH, indWiC, indWoC, indWkH, indOoH); std::string expectedGradOShape = ShapeUtils::shapeAsString( ShapeUtils::composeShapeUsingDimsAndIdx({bS, iC, oH, oW, 0, indIOioC, indIiH, indIiH + 1})); std::string expectedGradIShape = ShapeUtils::shapeAsString( ShapeUtils::composeShapeUsingDimsAndIdx({bS, iC, iH, iW, 0, indIOioC, indIiH, indIiH + 1})); REQUIRE_TRUE(expectedGradOShape == ShapeUtils::shapeAsString(gradO), 0, "AVGPOOL2D_BP op: wrong shape of output's gradients array (next epsilon), expected is %s, but got %s instead !", expectedGradOShape.c_str(), ShapeUtils::shapeAsString(gradO).c_str()); REQUIRE_TRUE(expectedGradIShape == ShapeUtils::shapeAsString(gradI), 0, "AVGPOOL2D_BP op: wrong shape of input's gradients array (epsilon), expected is %s, but got %s instead !", expectedGradIShape.c_str(), ShapeUtils::shapeAsString(gradI).c_str()); if (!isNCHW) { input = new NDArray(input->permute( {0, 3, 1, 2})); // [bS, iH, iW, iC] -> [bS, iC, iH, iW] gradI = new NDArray(gradI->permute( {0, 3, 1, 2})); // [bS, iH, iW, iC] -> [bS, iC, iH, iW] gradO = new NDArray(gradO->permute( {0, 3, 1, 2})); // [bS, oH, oW, iC] -> [bS, iC, oH, oW] } if (isSameMode) // SAME ConvolutionUtils::calcPadding2D(pH, pW, oH, oW, iH, iW, kH, kW, sH, sW, dH, dW); auto poolingMode = PoolingType::MAX_POOL; dnnl_memory_desc_t empty; dnnl::memory::desc pool_src_md(empty), pool_diff_src_md(empty), pool_dst_md(empty); dnnl::memory::desc user_src_md(empty), user_diff_src_md(empty), user_dst_md(empty); dnnl::memory::dims pool_strides, pool_kernel, pool_padding, pool_padding_r; dnnl::algorithm algorithm; mkldnnUtils::getMKLDNNMemoryDescPool2d(kH, kW, sH, sW, pH, pW, dH, dW, poolingMode, extraParam0, true, bS, iC, iH, iW, oC, oH, oW, input, gradI, gradO, algorithm, &pool_src_md, &pool_diff_src_md, &pool_dst_md, &user_src_md, &user_diff_src_md, &user_dst_md, pool_strides, pool_kernel, pool_padding, pool_padding_r); // input is sometimes null, so we can't rely on pool_src_md being valid auto pool_desc = pooling_forward::desc(prop_kind::forward, algorithm, input->buffer() != nullptr ? pool_src_md : pool_diff_src_md, pool_dst_md, pool_strides, pool_kernel, pool_padding, pool_padding_r); auto engine = mkldnnUtils::getEngine(LaunchContext::defaultContext()->engine()); dnnl::stream stream(engine); auto pool_prim_desc = pooling_forward::primitive_desc(pool_desc, engine); auto poolB_desc = pooling_backward::desc(algorithm, pool_diff_src_md, pool_dst_md, pool_strides, pool_kernel, pool_padding, pool_padding_r); auto poolB_prim_desc = pooling_backward::primitive_desc(poolB_desc, engine, pool_prim_desc); auto userB_src_memory = dnnl::memory(user_src_md, engine, gradI->buffer()); auto userB_dst_memory = dnnl::memory(user_dst_md, engine, gradO->buffer()); auto poolB_src_memory = userB_src_memory; if (poolB_prim_desc.diff_src_desc() != userB_src_memory.get_desc()) { poolB_src_memory = dnnl::memory(poolB_prim_desc.diff_src_desc(), engine); } auto poolB_dst_memory = userB_dst_memory; if (poolB_prim_desc.diff_dst_desc() != userB_dst_memory.get_desc()) { poolB_dst_memory = dnnl::memory(poolB_prim_desc.diff_dst_desc(), engine); reorder(userB_dst_memory, poolB_dst_memory).execute(stream, userB_dst_memory, poolB_dst_memory); } auto user_src_memory = dnnl::memory(user_src_md, engine, input->buffer()); auto pool_src_memory = user_src_memory; if (pool_prim_desc.src_desc() != user_src_memory.get_desc()) { pool_src_memory = dnnl::memory(pool_prim_desc.src_desc(), engine); reorder(user_src_memory, pool_src_memory).execute(stream, user_src_memory, pool_src_memory); } auto pool_dst_memory = dnnl::memory(pool_prim_desc.dst_desc(), engine); auto pool_workspace_memory = dnnl::memory(pool_prim_desc.workspace_desc(), engine); pooling_forward(pool_prim_desc).execute(stream, {{DNNL_ARG_SRC, pool_src_memory}, {DNNL_ARG_DST, pool_dst_memory}, {DNNL_ARG_WORKSPACE, pool_workspace_memory}}); // probably wrong, fix that pooling_backward(poolB_prim_desc).execute(stream, {{DNNL_ARG_DIFF_DST, poolB_dst_memory}, {DNNL_ARG_WORKSPACE, pool_workspace_memory}, {DNNL_ARG_DIFF_SRC, poolB_src_memory}}); if (poolB_prim_desc.diff_src_desc() != userB_src_memory.get_desc()) { reorder(poolB_src_memory, userB_src_memory).execute(stream, poolB_src_memory, userB_src_memory); } stream.wait(); if (!isNCHW) { delete input; delete gradI; delete gradO; } return Status::OK(); } PLATFORM_CHECK(maxpool2d_bp) { // we don't want to use mkldnn if cpu doesn't support avx/avx2 if (::optimalLevel() < 2) return false; auto input = INPUT_VARIABLE(0); auto output = OUTPUT_VARIABLE(0); return block.isUseMKLDNN() && nd4j::MKLDNNStream::isSupported({input, output}); } } } }