// ---------------------------------------------------------------------------- // Copyright 2017 Nervana Systems Inc. // 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 // ---------------------------------------------------------------------------- #pragma once #include <memory> #include "ngraph/log.hpp" #include "ngraph/types/element_type.hpp" #include "ngraph/util.hpp" #include "util/test_tools.hpp" namespace ngraph { class Node; class Function; namespace runtime { class Backend; class Manager; } // namespace runtime namespace autodiff { template <typename T> std::vector<std::shared_ptr<runtime::TensorView>> backprop_derivative(const std::shared_ptr<runtime::Manager>& manager, const std::shared_ptr<runtime::Backend>& backend, const std::shared_ptr<Function>& f, const std::vector<std::shared_ptr<runtime::TensorView>>& args, const std::vector<std::shared_ptr<op::Parameter>>& indep_params) { // y = f(X) // using X (upper case) to denote all paramenters of f // using x (lower case) to denote an individual paramemter of f a.k.a. Xj // NOTE: using X* to denote all x "of interest" represented by indep_params Shape y_shape = f->get_output_shape(0); // adjoint auto c_param = std::make_shared<op::Parameter>(element::from<T>(), y_shape); auto c_arg = backend->make_primary_tensor_view<T>(y_shape); // df/dX* // return value for f'(X, c) std::vector<std::shared_ptr<Node>> df_output_params; std::vector<std::shared_ptr<runtime::TensorView>> df_output_args; // return value for this function std::vector<std::shared_ptr<runtime::TensorView>> results; // for each x "of interest" for (auto x : indep_params) { auto x_shape = x->get_shape(); // each element of y has a derivative with respect to each element of x // hence, create a y by x sized tensor for this result auto y_by_x_shape = y_shape; y_by_x_shape.insert(y_by_x_shape.end(), x_shape.begin(), x_shape.end()); results.push_back(backend->make_primary_tensor_view<T>(y_by_x_shape)); // add df/dx to df/dX* df_output_params.push_back(f->get_output_op(0)->backprop_node(x, c_param)); df_output_args.push_back(backend->make_primary_tensor_view<T>(x_shape)); } // (X, c) // input to f'(X, c) std::vector<std::shared_ptr<op::Parameter>> df_input_params = f->get_parameters(); df_input_params.push_back(c_param); // df/dX* = f'(X, c) auto df = std::make_shared<Function>(df_output_params, df_input_params); // create fprop cache // creates modified forward function -> (y, cached) = f(x) // creates modified backward function -> df/dX* = f'(c, cached) auto fprop_cache = cache_fprop(f, df, {c_param}); // modified f outputs std::vector<std::shared_ptr<ngraph::runtime::TensorView>> f_output_args; f_output_args.push_back(backend->make_primary_tensor_view<T>(y_shape)); // modified f' inputs std::vector<std::shared_ptr<ngraph::runtime::TensorView>> df_input_args; df_input_args.push_back(c_arg); // add cached nodes to both modified f outputs and modified f' inputs for (auto node : fprop_cache.fprop_output_nodes) { auto tv = backend->make_primary_tensor_view<T>(node->get_shape()); df_input_args.push_back(tv); f_output_args.push_back(tv); } // compile and run modified (y, cached) = f(x) auto cache_fwd = manager->compile(fprop_cache.fprop); auto cache_fwd_cf = backend->make_call_frame(cache_fwd); cache_fwd_cf->tensor_call(args, f_output_args); // compile modified df/dX* = f'(c, cached) auto external = manager->compile(fprop_cache.bprop); auto cf = backend->make_call_frame(external); // create storage for results // * outer vector size = number of x "of interest" // * inner vector size = number of elements in y * number of elements in x std::vector<std::vector<T>> result_vect; std::vector<typename std::vector<T>::iterator> result_pos; for (auto result : results) { result_vect.push_back(read_vector<T>(result)); result_pos.push_back(result_vect.back().begin()); } // get adjoint and force to all elements to zero auto c_vec = read_vector<T>(c_arg); fill(c_vec.begin(), c_vec.end(), 0); // for each element of the adjoint // same as saying for each element of y for (size_t i = 0; i < c_vec.size(); i++) { // set a single adjoint element c_vec[i] = 1; write_vector(c_arg, c_vec); // call modified df/dX* = f'(c, cached) cf->tensor_call(df_input_args, df_output_args); // reset the adjoint element c_vec[i] = 0; write_vector(c_arg, c_vec); // for each result // same as saying for each x "of interest" for (size_t j = 0; j < results.size(); j++) { // copy df/dx to storage for this element of y auto dfdx = read_vector<T>(df_output_args[j]); result_pos[j] = std::copy(dfdx.begin(), dfdx.end(), result_pos[j]); } } // copy storage to results and return for (size_t j = 0; j < results.size(); j++) { write_vector(results[j], result_vect[j]); } return results; } } }