astar_readmodel.cpp

/*
    Copyright 2005-2016 Intel Corporation.  All Rights Reserved.

    This file is part of Threading Building Blocks. Threading Building Blocks is free software;
    you can redistribute it and/or modify it under the terms of the GNU General Public License
    version 2  as  published  by  the  Free Software Foundation.  Threading Building Blocks is
    distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the
    implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
    See  the GNU General Public License for more details.   You should have received a copy of
    the  GNU General Public License along with Threading Building Blocks; if not, write to the
    Free Software Foundation, Inc.,  51 Franklin St,  Fifth Floor,  Boston,  MA 02110-1301 USA

    As a special exception,  you may use this file  as part of a free software library without
    restriction.  Specifically,  if other files instantiate templates  or use macros or inline
    functions from this file, or you compile this file and link it with other files to produce
    an executable,  this file does not by itself cause the resulting executable to be covered
    by the GNU General Public License. This exception does not however invalidate any other
    reasons why the executable file might be covered by the GNU General Public License.
*/

#include <cstdio>
#include <vector>
#include <map>
#include <math.h>
#include <fstream>
#include <iostream>
#include <sstream>
#include <string>
//#include <armadillo>

#include "tbb/atomic.h"
#include "tbb/tick_count.h"
#include "tbb/task_scheduler_init.h"
#include "tbb/task_group.h"
#include "tbb/concurrent_priority_queue.h"
#include "tbb/spin_mutex.h"
#include "tbb/parallel_for.h"
#include "tbb/blocked_range.h"
#include "../../common/utility/utility.h"
#include "../../common/utility/fast_random.h"

#if defined(_MSC_VER) && defined(_Wp64)
    // Workaround for overzealous compiler warnings in /Wp64 mode
    #pragma warning (disable: 4267)
#endif /* _MSC_VER && _Wp64 */

using namespace std;
using namespace tbb;

struct point {
    double x, y;
    point() {}
    point(double _x, double _y) : x(_x), y(_y) {}
    point(const point& p) : x(p.x), y(p.y) {}
};

double get_distance(const point& p1, const point& p2) {
    double xdiff=p1.x-p2.x, ydiff=p1.y-p2.y;
    return sqrt(xdiff*xdiff + ydiff*ydiff);
}

// generates random points on 2D plane within a box of maxsize width & height
point generate_random_point(utility::FastRandom& mr) {
    const size_t maxsize=500;
    double x = (double)(mr.get() % maxsize);
    double y = (double)(mr.get() % maxsize);
    return point(x,y);
}

// weighted toss makes closer nodes (in the point vector) heavily connected
bool die_toss(size_t a, size_t b, utility::FastRandom& mr) {
    int node_diff = std::abs((int)(a-b));
    // near nodes
    if (node_diff < 16) return true;
    // mid nodes
    if (node_diff < 64) return ((int)mr.get() % 8 == 0);
    // far nodes
    if (node_diff < 512) return ((int)mr.get() % 16 == 0);
    return false;
}

typedef vector<point> point_set;
typedef size_t vertex_id;
typedef std::pair<vertex_id,double> vertex_rec;
typedef vector<vector<vertex_id> > edge_set;

bool verbose = false;          // prints bin details and other diagnostics to screen
bool silent = false;           // suppress all output except for time
size_t N = 1000;               // number of vertices
size_t src = 0;                // start of path
size_t dst = N-1;              // end of path
double INF=100000.0;           // infinity
size_t grainsize = 16;         // number of vertices per task on average
size_t max_spawn;              // max tasks to spawn
tbb::atomic<size_t> num_spawn;      // number of active tasks

point_set vertices;            // vertices
edge_set edges;                // edges
vector<vertex_id> predecessor; // for recreating path from src to dst

vector<double> f_distance;     // estimated distances at particular vertex
vector<double> g_distance;     // current shortest distances from src vertex
spin_mutex    *locks;          // a lock for each vertex
task_group *sp_group;          // task group for tasks executing sub-problems

class compare_f {
public:
    bool operator()(const vertex_rec& u, const vertex_rec& v) const {
        return u.second>v.second;
    }
};

concurrent_priority_queue<vertex_rec, compare_f> open_set; // tentative vertices

void shortpath_helper();

#if !__TBB_CPP11_LAMBDAS_PRESENT
class shortpath_helper_functor {
public:
    shortpath_helper_functor() {};
    void operator() () const { shortpath_helper(); }
};
#endif

void shortpath() {
    sp_group = new task_group;
    g_distance[src] = 0.0; // src's distance from src is zero
    f_distance[src] = get_distance(vertices[src], vertices[dst]); // estimate distance from src to dst
    open_set.push(make_pair(src,f_distance[src])); // push src into open_set
#if __TBB_CPP11_LAMBDAS_PRESENT
    sp_group->run([](){ shortpath_helper(); });
#else
    sp_group->run( shortpath_helper_functor() );
#endif
    sp_group->wait();
    delete sp_group;
}

void shortpath_helper() {
    vertex_rec u_rec;
    while (open_set.try_pop(u_rec)) {
        vertex_id u = u_rec.first;
        if (u==dst) continue;
        double f = u_rec.second;
        double old_g_u = 0.0;
        {
            spin_mutex::scoped_lock l(locks[u]);
            if (f > f_distance[u]) continue; // prune search space
            old_g_u = g_distance[u];
        }
        for (size_t i=0; i<edges[u].size(); ++i) {
            vertex_id v = edges[u][i];
            double new_g_v = old_g_u + get_distance(vertices[u], vertices[v]);
            double new_f_v = 0.0;
            // the push flag lets us move some work out of the critical section below
            bool push = false;
            {
                spin_mutex::scoped_lock l(locks[v]);
                if (new_g_v < g_distance[v]) {
                    predecessor[v] = u;
                    g_distance[v] = new_g_v;
                    new_f_v = f_distance[v] = g_distance[v] + get_distance(vertices[v], vertices[dst]);
                    push = true;
                }
            }
            if (push) {
                open_set.push(make_pair(v,new_f_v));
                size_t n_spawn = ++num_spawn;
                if (n_spawn < max_spawn) {
#if __TBB_CPP11_LAMBDAS_PRESENT
                    sp_group->run([]{ shortpath_helper(); });
#else
                    sp_group->run( shortpath_helper_functor() );
#endif
                }
                else --num_spawn;
            }
        }
    }
    --num_spawn;
}

void make_path(vertex_id src, vertex_id dst, vector<vertex_id>& path) {
    vertex_id at = predecessor[dst];
    if (at == N) path.push_back(src);
    else if (at == src) { path.push_back(src); path.push_back(dst); }
    else { make_path(src, at, path); path.push_back(dst); }
}

void print_path() {
    vector<vertex_id> path;
    double path_length=0.0;
    make_path(src, dst, path);
    if (verbose) printf("\n      ");
    for (size_t i=0; i<path.size(); ++i) {
        if (path[i] != dst) {
            double seg_length = get_distance(vertices[path[i]], vertices[path[i+1]]);
            if (verbose) printf("%6.1f       ", seg_length);
            path_length += seg_length;
        }
        else if (verbose) printf("\n");
    }
    if (verbose) {
        for (size_t i=0; i<path.size(); ++i) {
            if (path[i] != dst) printf("(%4d)------>", (int)path[i]);
            else printf("(%4d)\n", (int)path[i]);
        }
    }
    if (verbose) printf("Total distance = %5.1f\n", path_length);
    else if (!silent) printf(" %5.1f\n", path_length);
}

int get_default_num_threads() {
    static int threads = 0;
    if (threads == 0)
        threads = tbb::task_scheduler_init::default_num_threads();
    return threads;
}

#if !__TBB_CPP11_LAMBDAS_PRESENT
class gen_vertices {
public:
    gen_vertices() {}
    void operator() (blocked_range<size_t>& r) const {
        utility::FastRandom my_random((unsigned int)r.begin());
        for (size_t i=r.begin(); i!=r.end(); ++i) {
            vertices[i] = generate_random_point(my_random);
        }
    }
};

class gen_edges {
public:
    gen_edges() {}
    void operator() (blocked_range<size_t>& r) const {
        utility::FastRandom my_random((unsigned int)r.begin());
        for (size_t i=r.begin(); i!=r.end(); ++i) {
            for (size_t j=0; j<i; ++j) {
                if (die_toss(i, j, my_random))
                    edges[i].push_back(j);
            }
        }
    }
};

class reset_vertices {
public:
    reset_vertices() {}
    void operator() (blocked_range<size_t>& r) const {
        for (size_t i=r.begin(); i!=r.end(); ++i) {
            f_distance[i] = g_distance[i] = INF;
            predecessor[i] = N;
        }
    }
};
#endif

void InitializeGraph() {
    task_scheduler_init init(get_default_num_threads());
    vertices.resize(N);
    edges.resize(N);
    predecessor.resize(N);
    g_distance.resize(N);
    f_distance.resize(N);
    locks = new spin_mutex[N];
    if (verbose) printf("Generating vertices...\n");
#if __TBB_CPP11_LAMBDAS_PRESENT
    parallel_for(blocked_range<size_t>(0,N,64),
                 [&](blocked_range<size_t>& r) {
                     utility::FastRandom my_random(r.begin());
                     for (size_t i=r.begin(); i!=r.end(); ++i) {
                         vertices[i] = generate_random_point(my_random);
                     }
                 }, simple_partitioner());
#else
    parallel_for(blocked_range<size_t>(0,N,64), gen_vertices(), simple_partitioner());
#endif
    if (verbose) printf("Generating edges...\n");
#if __TBB_CPP11_LAMBDAS_PRESENT
    parallel_for(blocked_range<size_t>(0,N,64),
                 [&](blocked_range<size_t>& r) {
                     utility::FastRandom my_random(r.begin());
                     for (size_t i=r.begin(); i!=r.end(); ++i) {
                         for (size_t j=0; j<i; ++j) {
                             if (die_toss(i, j, my_random))
                                 edges[i].push_back(j);
                         }
                     }
                 }, simple_partitioner());
#else
    parallel_for(blocked_range<size_t>(0,N,64), gen_edges(), simple_partitioner());
#endif
    for (size_t i=0; i<N; ++i) {
        for (size_t j=0; j<edges[i].size(); ++j) {
            vertex_id k = edges[i][j];
            edges[k].push_back(i);
        }
    }
    if (verbose) printf("Done.\n");
}

void ReleaseGraph() {
    delete []locks;
}

void ResetGraph() {
    task_scheduler_init init(get_default_num_threads());
#if __TBB_CPP11_LAMBDAS_PRESENT
    parallel_for(blocked_range<size_t>(0,N),
                 [&](blocked_range<size_t>& r) {
                     for (size_t i=r.begin(); i!=r.end(); ++i) {
                         f_distance[i] = g_distance[i] = INF;
                         predecessor[i] = N;
                     }
                 });
#else
    parallel_for(blocked_range<size_t>(0,N), reset_vertices());
#endif
}

int main(int argc, char *argv[]) {
    try {
        utility::thread_number_range threads(get_default_num_threads);
        utility::parse_cli_arguments(argc, argv,
                                     utility::cli_argument_pack()
                                     //"-h" option for displaying help is present implicitly
                                     .positional_arg(threads,"threads",utility::thread_number_range_desc)
                                     .arg(verbose,"verbose","   print diagnostic output to screen")
                                     .arg(silent,"silent","    limits output to timing info; overrides verbose")
                                     .arg(N,"N","         number of vertices")
                                     .arg(src,"start","      start of path")
                                     .arg(dst,"end","        end of path")
        );
        
        if (silent) verbose = false;  // make silent override verbose
        else
            printf("shortpath will run with %d vertices to find shortest path between vertices"
                   " %d and %d using %d:%d threads.\n",
                   (int)N, (int)src, (int)dst, (int)threads.first, (int)threads.last);

        if (dst >= N) {
            if (verbose)
                printf("end value %d is invalid for %d vertices; correcting to %d\n", (int)dst, (int)N, (int)N-1);
            dst = N-1;
        }

        num_spawn = 0;
        max_spawn = N/grainsize;
        tick_count t0, t1;
        InitializeGraph();
        for (int n_thr=threads.first; n_thr<=threads.last; n_thr=threads.step(n_thr)) {
            ResetGraph();
            task_scheduler_init init(n_thr);
            t0 = tick_count::now();
            shortpath();
            t1 = tick_count::now();
            if (!silent) {
                if (predecessor[dst] != N) {
                    printf("%d threads: [%6.6f] The shortest path from vertex %d to vertex %d is:",
                           (int)n_thr, (t1-t0).seconds(), (int)src, (int)dst);
                    print_path();
                }
                else {
                    printf("%d threads: [%6.6f] There is no path from vertex %d to vertex %d\n",
                           (int)n_thr, (t1-t0).seconds(), (int)src, (int)dst);
                }
            } else
                utility::report_elapsed_time((t1-t0).seconds());
        }
        ReleaseGraph();
        return 0;
    } catch(std::exception& e) {
        cerr<<"error occurred. error text is :\"" << e.what() << "\"\n";
        return 1;
    }
}

#define contextWindow 3
#define wspan  7

map< string, vector<double> > loadDoubleMap(string filename) {
    map< string, vector<double> > matrix;
    ifstream infile(filename);
    
    while (infile) {
        string s;
        if (!getline(infile, s)) break;
        
        istringstream ss(s);
        vector <double> row;
        string name;
        getline(ss, name, ' ');
        while (ss) {
            string s;
            if (!getline(ss, s, ' ')) break;
            row.push_back(stod(s));
        }
        matrix[name] = row;
    }
    
    return matrix;
}

vector< vector<double> > loadDoubleMatrix(string filename) {
    vector< vector<double> > matrix;
    ifstream infile(filename);
    
    while (infile) {
        string s;
        if (!getline(infile, s)) break;
        
        istringstream ss(s);
        vector <double> row;
        
        while (ss) {
            string s;
            if (!getline(ss, s, ' ')) break;
            row.push_back(stod(s));
        }
        
        matrix.push_back(row);
    }
    
    return matrix;
}


#define leftPad           "*left_pad*"
#define rightPad          "*right_pad*"
#define unknownLower      "*unknown_lower*"
#define unknownUpper      "*unknown_upper*"
#define unknownSpecial    "*unknown_special*"
#define capsLower         "*lower_case*"
#define capsUpper         "*upper_case*"
#define capitalizedPad    "*caps_pad*"
#define suffixPad         "*suffix_pad*"
#define unknownSuffix     "*unknown_suffix*"

void loadModel() {
    string modelFolder = "";
    auto embeddings = loadDoubleMap(modelFolder + "embeddings");
    auto capitals = loadDoubleMap(modelFolder + "capitals");
    auto suffix = loadDoubleMap(modelFolder + "suffix");
    
    int totalFeatures =
    (embeddings[unknownLower].size() + suffix[unknownSuffix].size() +
     capitals[capsLower].size()) * (2 * contextWindow + 1);
    
    //auto lexicalCategories = loadCategories("categories");
    //load embeddings
    auto weightMatrix = loadDoubleMatrix("classifier");
    auto tagDict = loadDoubleMap("tagdict");
}