MST.c

//  Copyright 2013 Google Inc. All Rights Reserved.
//
//  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.

// ./word2vec -train /data/disk1/private/nyl/copus2.txt -output vectors12.bin -cbow 0 -size 200 -window 8 -negative 25 -hs 0 -sample 1e-4 -threads 30 -binary 1 -iter 1 -read-vocab ../data2/ReadVocab2700000000 -read-meaning ../ReadMeaning -read-sense ../data2/ReadSenseWord2700000000 -min-count 1

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <pthread.h>

#define MAX_STRING 100
#define EXP_TABLE_SIZE 1000
#define EXP_FROM_ZERO_FORE 1000
#define MAX_EXP 6
#define MAX_SENTENCE_LENGTH 1000
#define MAX_CODE_LENGTH 40

const int vocab_hash_size = 10000000;  // Maximum 30 * 0.7 = 21M words in the vocabulary
const int meaning_hash_size = 10000000;  // Maximum 30 * 0.7 = 21M words in the vocabulary

unsigned long long next_random = 19960322; 

typedef float real;                    // Precision of float numbers

real pre_exp[EXP_TABLE_SIZE];

struct vocab_word {
  long long cn;
  int *point;
  char *word, *code, codelen;
};

struct meaning_word {
  char *word;
};

real vectorDot(real *a, real *b, int l) {
  int i;
  real dot = 0;
  for (i = 0; i < l; ++i)
    dot += a[i] * b[i];
  return dot;
}

real *meaning_syn;

struct sense {
  int num;
  real *mult_sense_value;
  int *meaning_word_cnt;
  int *meaning_word_rank;
};

  void init(struct sense *s, int n, int l, int *rank, int *cnt) {
    int i;
    s->num = n;
    s->mult_sense_value = (real *)calloc(n * l, sizeof(real));
    for (i = 0; i < n * l; ++i) {
      next_random = next_random * (unsigned long long)25214903917 + 11;
      s->mult_sense_value[i] = (((next_random & 0xFFFF) / (real)65536) - 0.5) / l;
    }
    if (n > 1) {
      s->meaning_word_cnt = (int *)calloc(n, sizeof(int));
      for (i = 0; i < n; ++i)
        s->meaning_word_cnt[i] = cnt[i];
      s->meaning_word_rank = (int *)calloc(s->meaning_word_cnt[n - 1], sizeof(int));
      for (i = 0; i < s->meaning_word_cnt[n - 1]; ++i)
        s->meaning_word_rank[i] = rank[i];
    }
  }
  real meaning_word_dot(struct sense *s, int r, real *b, int l, real *a) {
    int i, j;
    for (i = 0; i < l; ++i) a[i] = 0;
    for (i = (r == 0 ? 0 : s->meaning_word_cnt[r - 1]); i < s->meaning_word_cnt[r]; ++i) {
      real *c = &meaning_syn[s->meaning_word_rank[i] * l];
      for (j = 0; j < l; ++j) a[j] += c[j];
    }
    real ans = vectorDot(a, b, l);
    return ans;
  }

  void update(struct sense *s, real *theta, real *attention, real *_exp, int l, real *delta, real total) {
    int i, j, k;
    if (s->num == 1) {
      for (i = 0; i < l; ++i)
        s->mult_sense_value[i] += delta[i];
      return;
    }
    for (i = 0; i < s->num; ++i)
      for (j = (i == 0 ? 0 : s->meaning_word_cnt[i - 1]); j < s->meaning_word_cnt[i]; ++j)
        for (k = 0; k < l; ++k)
          meaning_syn[s->meaning_word_rank[j] * l + k] += s->mult_sense_value[i * l + k] * theta[k] * (1 - _exp[i]) * _exp[i] * total;
    for (i = 0; i < s->num; ++i)
      for (j = 0; j < l; ++j)
        s->mult_sense_value[i * l + j] += delta[j] * _exp[i];
  }

char train_file[MAX_STRING], output_file[MAX_STRING], checkpoint[MAX_STRING];
char save_vocab_file[MAX_STRING], read_vocab_file[MAX_STRING], read_meaning_file[MAX_STRING], read_sense_file[MAX_STRING];
struct vocab_word *vocab;
struct meaning_word *meaning;
int binary = 0, cbow = 1, debug_mode = 2, window = 5, min_count = 50, num_threads = 12, min_reduce = 1;
int *vocab_hash, *meaning_hash;
long long vocab_max_size = 1000, vocab_size = 0, layer1_size = 100;
long long meaning_size = 0;
long long train_words = 0, word_count_actual = 0, iter = 5, file_size = 0, classes = 0;
real alpha = 0.00375, starting_alpha, sample = 1e-3;
real *syn1, *syn1neg, *expTable;
struct sense *syn0;
clock_t start;

int hs = 0, negative = 5;
const int table_size = 1e8;
int *table;

void InitUnigramTable() {
  int a, i;
  double train_words_pow = 0;
  double d1, power = 0.75;
  table = (int *)malloc(table_size * sizeof(int));
  for (a = 0; a < vocab_size; a++) train_words_pow += pow(vocab[a].cn, power);
  i = 0;
  d1 = pow(vocab[i].cn, power) / train_words_pow;
  for (a = 0; a < table_size; a++) {
    table[a] = i;
    if (a / (double)table_size > d1) {
      i++;
      d1 += pow(vocab[i].cn, power) / train_words_pow;
    }
    if (i >= vocab_size) i = vocab_size - 1;
  }
}

int ReadVocabInt(FILE *fin) {
  int a = 0, ch, sub = '0';
  while (!feof(fin)) {
    ch = fgetc(fin);
    if ((ch == ' ') || (ch == '\n'))
      break;
    a = a * 10 + ch - sub;
  }
  return a;
}

void ReadVocabWord(char *word, FILE *fin) {
  int a = 0, ch;
  while (!feof(fin)) {
    ch = fgetc(fin);
    if ((ch == ' ') || (ch == '\n'))
      break;
    word[a] = ch;
    ++a;
    if (a >= MAX_STRING - 1)
      --a;
  }
  word[a] = 0;
}

void ReadSenseWord(char *word, FILE *fin, int *rank, int *cnt) {
  int i, j;
  ReadVocabWord(word, fin);
  int loc = SearchVocab(word);
  int num = ReadVocabInt(fin);
  if (num == 1) {
    init(&syn0[loc], num, layer1_size, rank, cnt);
    return;
  }
  for (i = 0; i < num; ++i) {
    int meaning_cnt = ReadVocabInt(fin);
    cnt[i] = (i == 0 ? 0 : cnt[i - 1]) + meaning_cnt;
    for (j = (i == 0 ? 0 : cnt[i - 1]); j < cnt[i]; ++j) {
      ReadVocabWord(word, fin);
      rank[j] = SearchMeaning(word);
    }
  }
  init(&syn0[loc], num, layer1_size, rank, cnt);
}

// Reads a single word from a file, assuming space + tab + EOL to be word boundaries
void ReadWord(char *word, FILE *fin) {
  int a = 0, ch;
  while (!feof(fin)) {
    ch = fgetc(fin);
    if (ch == 13) continue;
    if ((ch == ' ') || (ch == '\t') || (ch == '\n')) {
      if (a > 0) {
        if (ch == '\n') ungetc(ch, fin);
        break;
      }
      if (ch == '\n') {
        strcpy(word, (char *)"</s>");
        return;
      } else continue;
    }
    word[a] = ch;
    a++;
    if (a >= MAX_STRING - 1) a--;   // Truncate too long words
  }
  word[a] = 0;
}

// Reads a single word from a file, assuming space + tab + EOL to be word boundaries
void ReadMeaningWord(char *word, FILE *fin) {
  int a = 0, ch;
  while (!feof(fin)) {
    ch = fgetc(fin);
    if ((ch == ' ') || (ch == '\n'))
      break;
    word[a] = ch;
    ++a;
    if (a >= MAX_STRING - 1)
      --a;
  }
  word[a] = 0;
}

// Returns hash value of a word
int GetWordHash(char *word) {
  unsigned long long a, hash = 0;
  for (a = 0; a < strlen(word); a++) hash = hash * 257 + word[a];
  hash = hash % vocab_hash_size;
  return hash;
}

// Returns position of a word in the vocabulary; if the word is not found, returns -1
int SearchVocab(char *word) {
  unsigned int hash = GetWordHash(word);
  while (1) {
    if (vocab_hash[hash] == -1) return -1;
    if (!strcmp(word, vocab[vocab_hash[hash]].word)) return vocab_hash[hash];
    hash = (hash + 1) % vocab_hash_size;
  }
  return -1;
}

// Returns position of a word in the meaning vocabulary; if the word is not found, returns -1
int SearchMeaning(char *word) {
  unsigned int hash = GetWordHash(word);
  while (1) {
    if (meaning_hash[hash] == -1) return -1;
    if (!strcmp(word, meaning[meaning_hash[hash]].word)) return meaning_hash[hash];
    hash = (hash + 1) % meaning_hash_size;
  }
  return -1;
}

// Reads a word and returns its index in the vocabulary
int ReadWordIndex(FILE *fin) {
  char word[MAX_STRING];
  ReadWord(word, fin);
  if (feof(fin)) return -1;
  return SearchVocab(word);
}

// Adds a word to the vocabulary
int AddWordToVocab(char *word) {
  unsigned int hash, length = strlen(word) + 1;
  if (length > MAX_STRING) length = MAX_STRING;
  vocab[vocab_size].word = (char *)calloc(length, sizeof(char));
  strcpy(vocab[vocab_size].word, word);
  vocab[vocab_size].cn = 0;
  vocab_size++;
  // Reallocate memory if needed
  if (vocab_size + 2 >= vocab_max_size) {
    vocab_max_size += 1000;
    vocab = (struct vocab_word *)realloc(vocab, vocab_max_size * sizeof(struct vocab_word));
  }
  hash = GetWordHash(word);
  while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size;
  vocab_hash[hash] = vocab_size - 1;
  return vocab_size - 1;
}

// Adds a meaning to the meaning vocabulary
int AddWordToMeaningVocab(char *word) {
  unsigned int hash, length = strlen(word) + 1;
  if (length > MAX_STRING) length = MAX_STRING;
  meaning[meaning_size].word = (char *)calloc(length, sizeof(char));
  strcpy(meaning[meaning_size].word, word);
  meaning_size++;
  hash = GetWordHash(word);
  while (meaning_hash[hash] != -1) hash = (hash + 1) % meaning_hash_size;
  meaning_hash[hash] = meaning_size - 1;
  return meaning_size - 1;
}

// Used later for sorting by word counts
int VocabCompare(const void *a, const void *b) {
    return ((struct vocab_word *)b)->cn - ((struct vocab_word *)a)->cn;
}

// Sorts the vocabulary by frequency using word counts
void SortVocab() {
  int a, size;
  unsigned int hash;
  // Sort the vocabulary and keep </s> at the first position
  qsort(&vocab[0], vocab_size - 1, sizeof(struct vocab_word), VocabCompare);
  for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1;
  size = vocab_size;
  train_words = 0;
  for (a = 0; a < size; a++) {
    // Words occuring less than min_count times will be discarded from the vocab
    if ((vocab[a].cn < min_count) && (a != 0)) {
      vocab_size--;
      free(vocab[a].word);
    } else {
      // Hash will be re-computed, as after the sorting it is not actual
      hash=GetWordHash(vocab[a].word);
      while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size;
      vocab_hash[hash] = a;
      train_words += vocab[a].cn;
    }
  }
  vocab = (struct vocab_word *)realloc(vocab, (vocab_size + 1) * sizeof(struct vocab_word));
  // Allocate memory for the binary tree construction
  for (a = 0; a < vocab_size; a++) {
    vocab[a].code = (char *)calloc(MAX_CODE_LENGTH, sizeof(char));
    vocab[a].point = (int *)calloc(MAX_CODE_LENGTH, sizeof(int));
  }
}

// Reduces the vocabulary by removing infrequent tokens
void ReduceVocab() {
  int a, b = 0;
  unsigned int hash;
  for (a = 0; a < vocab_size; a++) if (vocab[a].cn > min_reduce) {
    vocab[b].cn = vocab[a].cn;
    vocab[b].word = vocab[a].word;
    b++;
  } else free(vocab[a].word);
  vocab_size = b;
  for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1;
  for (a = 0; a < vocab_size; a++) {
    // Hash will be re-computed, as it is not actual
    hash = GetWordHash(vocab[a].word);
    while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size;
    vocab_hash[hash] = a;
  }
  fflush(stdout);
  min_reduce++;
}

// Create binary Huffman tree using the word counts
// Frequent words will have short uniqe binary codes
void CreateBinaryTree() {
  long long a, b, i, min1i, min2i, pos1, pos2, point[MAX_CODE_LENGTH];
  char code[MAX_CODE_LENGTH];
  long long *count = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long));
  long long *binary = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long));
  long long *parent_node = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long));
  for (a = 0; a < vocab_size; a++) count[a] = vocab[a].cn;
  for (a = vocab_size; a < vocab_size * 2; a++) count[a] = 1e15;
  pos1 = vocab_size - 1;
  pos2 = vocab_size;
  // Following algorithm constructs the Huffman tree by adding one node at a time
  for (a = 0; a < vocab_size - 1; a++) {
    // First, find two smallest nodes 'min1, min2'
    if (pos1 >= 0) {
      if (count[pos1] < count[pos2]) {
        min1i = pos1;
        pos1--;
      } else {
        min1i = pos2;
        pos2++;
      }
    } else {
      min1i = pos2;
      pos2++;
    }
    if (pos1 >= 0) {
      if (count[pos1] < count[pos2]) {
        min2i = pos1;
        pos1--;
      } else {
        min2i = pos2;
        pos2++;
      }
    } else {
      min2i = pos2;
      pos2++;
    }
    count[vocab_size + a] = count[min1i] + count[min2i];
    parent_node[min1i] = vocab_size + a;
    parent_node[min2i] = vocab_size + a;
    binary[min2i] = 1;
  }
  // Now assign binary code to each vocabulary word
  for (a = 0; a < vocab_size; a++) {
    b = a;
    i = 0;
    while (1) {
      code[i] = binary[b];
      point[i] = b;
      i++;
      b = parent_node[b];
      if (b == vocab_size * 2 - 2) break;
    }
    vocab[a].codelen = i;
    vocab[a].point[0] = vocab_size - 2;
    for (b = 0; b < i; b++) {
      vocab[a].code[i - b - 1] = code[b];
      vocab[a].point[i - b] = point[b] - vocab_size;
    }
  }
  free(count);
  free(binary);
  free(parent_node);
}

void LearnVocabFromTrainFile() {
  char word[MAX_STRING];
  FILE *fin;
  long long a, i;
  for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1;
  fin = fopen(train_file, "rb");
  if (fin == NULL) {
    printf("ERROR: training data file not found!\n");
    exit(1);
  }
  vocab_size = 0;
  AddWordToVocab((char *)"</s>");
  while (1) {
    ReadWord(word, fin);
    if (feof(fin)) break;
    train_words++;
    if ((debug_mode > 1) && (train_words % 100000 == 0)) {
      printf("%lldK%c", train_words / 1000, 13);
      fflush(stdout);
    }
    i = SearchVocab(word);
    if (i == -1) {
      a = AddWordToVocab(word);
      vocab[a].cn = 1;
    } else vocab[i].cn++;
    if (vocab_size > vocab_hash_size * 0.7) ReduceVocab();
  }
  SortVocab();
  if (debug_mode > 0) {
    printf("Vocab size: %lld\n", vocab_size);
    printf("Words in train file: %lld\n", train_words);
  }
  file_size = ftell(fin);
  fclose(fin);
}

void SaveVocab() {
  long long i;
  FILE *fo = fopen(save_vocab_file, "wb");
  for (i = 0; i < vocab_size; i++) fprintf(fo, "%s %lld\n", vocab[i].word, vocab[i].cn);
  fclose(fo);
}

void ReadVocab() {
  long long a, i = 0;
  char c;
  char word[MAX_STRING];
  FILE *fin = fopen(read_vocab_file, "rb");
  if (fin == NULL) {
    printf("Vocabulary file not found\n");
    exit(1);
  }
  for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1;
  vocab_size = 0;
  while (1) {
    ReadWord(word, fin);
    if (feof(fin)) break;
    a = AddWordToVocab(word);
    fscanf(fin, "%lld%c", &vocab[a].cn, &c);
    if (vocab[a].cn < min_count)
    	break;
    i++;
  }
  SortVocab();
  if (debug_mode > 0) {
    printf("Vocab size: %lld\n", vocab_size);
    printf("Words in train file: %lld\n", train_words);
  }
  fin = fopen(train_file, "rb");
  if (fin == NULL) {
    printf("ERROR: training data file not found!\n");
    exit(1);
  }
  fseek(fin, 0, SEEK_END);
  file_size = ftell(fin);
  fclose(fin);
  printf("%d\n", vocab_size);
}

void ReadPoint() {
	int a, b;
	FILE *fin = fopen(checkpoint, "r");
	fscanf(fin, "%lld %lld %lld %f\n", &meaning_size, &vocab_size, &layer1_size, &alpha);
	char c;
	char waste[MAX_STRING];
	for (a = 0; a < vocab_size; ++a) {
		fscanf(fin, "%s ", waste);
		fscanf(fin, "%d ", &(syn0[a].num));
		for (b = 0; b < syn0[a].num * layer1_size; ++b)
			fscanf(fin, "%f ", &(syn0[a].mult_sense_value[b]));
		if (syn0[a].num > 1) {
			for (b = 0; b < syn0[a].num; ++b)
				fscanf(fin, "%d ", &(syn0[a].meaning_word_cnt[b]));
			for (b = 0; b < syn0[a].meaning_word_cnt[syn0[a].num - 1]; ++b)
				fscanf(fin, "%d ", &(syn0[a].meaning_word_rank[b]));
		}
		fscanf(fin, "%c", &c);
	}
	for (a = 0; a < meaning_size; ++a) {
		fscanf(fin, "%s ", waste);
		for (b = 0; b < layer1_size; ++b)
			fscanf(fin, "%f ", &(meaning_syn[a * layer1_size + b]));
		fscanf(fin, "%c", &c);
	}
	for (a = 0; a < vocab_size * layer1_size; ++a)
		fscanf(fin, "%f ", &(syn1neg[a]));
	fscanf(fin, "%c", &c);
	fclose(fin);
	printf("checkpoint end\n");
}

void ReadMeaning() {
  long long a, i = 0;
  char c;
  char word[MAX_STRING];
  FILE *fin = fopen(read_meaning_file, "rb");
  if (fin == NULL) {
    printf("MeaningVocabulary file not found\n");
    exit(1);
  }
  for (a = 0; a < meaning_hash_size; a++) meaning_hash[a] = -1;
  meaning_size = 0;
  while (1) {
    ReadMeaningWord(word, fin);
    if (feof(fin)) break;
    a = AddWordToMeaningVocab(word);
  }
  printf("%d\n", meaning_size);
  fclose(fin);
}

void InitNet() {
  long long a, b;
  // unsigned long long next_random = 1;
  a = posix_memalign((void **)&syn0, 128, (long long)vocab_size * sizeof(struct sense));
  if (syn0 == NULL) {printf("Memory allocation failed\n"); exit(1);}
  a = posix_memalign((void **)&meaning_syn, 128, (long long)meaning_size * layer1_size * sizeof(real));
  if (meaning_syn == NULL) {printf("Memory allocation failed\n"); exit(1);}
  if (hs) {
    a = posix_memalign((void **)&syn1, 128, (long long)vocab_size * layer1_size * sizeof(real));
    if (syn1 == NULL) {printf("Memory allocation failed\n"); exit(1);}
    for (a = 0; a < vocab_size; a++) for (b = 0; b < layer1_size; b++)
     syn1[a * layer1_size + b] = 0;
  }
  if (negative>0) {
    a = posix_memalign((void **)&syn1neg, 128, (long long)vocab_size * layer1_size * sizeof(real));
    if (syn1neg == NULL) {printf("Memory allocation failed\n"); exit(1);}
    for (a = 0; a < vocab_size; a++) for (b = 0; b < layer1_size; b++)
     syn1neg[a * layer1_size + b] = 0;
  }
  for (a = 0; a < meaning_size; a++) for (b = 0; b < layer1_size; b++) {
    next_random = next_random * (unsigned long long)25214903917 + 11;
    meaning_syn[a * layer1_size + b] = (((next_random & 0xFFFF) / (real)65536) - 0.5) / layer1_size;
  }
  CreateBinaryTree();
  char *word = (char *)calloc(MAX_STRING, sizeof(char));
  int *rank = (int *)calloc(111, sizeof(int));
  int *cnt = (int *)calloc(111, sizeof(int));
  FILE *fi = fopen(read_sense_file, "r");
  for (a = 0; a < vocab_size; a++) {
    ReadSenseWord(word, fi, rank, cnt);
  }
  free(word);
  free(rank);
  free(cnt);
  fclose(fi);
}

void *TrainModelThread(void *id) {
  long long a, b, d, cw, word, last_word, sentence_length = 0, sentence_position = 0;
  int p, q;
  long long word_count = 0, last_word_count = 0, sen[MAX_SENTENCE_LENGTH + 1];
  long long l1, l2, c, target, label, local_iter = iter;
  // unsigned long long next_random = (long long)id;
  real f, g;
  clock_t now;
  real *neu1 = (real *)calloc(layer1_size, sizeof(real));
  real *neu1e = (real *)calloc(layer1_size, sizeof(real));
  FILE *fi = fopen(train_file, "rb");
  fseek(fi, file_size / (long long)num_threads * (long long)id, SEEK_SET);

  real *attention = (real *)calloc(layer1_size, sizeof(real));
  real *attention_syn = (real *)calloc(layer1_size, sizeof(real));
  real *_exp = (real *)calloc(223, sizeof(real));
  real *mult_part = (real *)calloc(layer1_size, sizeof(real));
  real *mult_part2 = (real *)calloc(layer1_size, sizeof(real));
  real *avg_meaning = (real *)calloc(layer1_size * 35, sizeof(real));
  real *avg_meaning_len = (real *)calloc(35, sizeof(real));
  real total = 0;

  while (1) {
    if (word_count - last_word_count > 10000) {
      word_count_actual += word_count - last_word_count;
      last_word_count = word_count;
      if ((debug_mode > 1)) {
        now=clock();
        printf("%cAlpha: %f  Progress: %.2f%%  Words/thread/sec: %.2fk  ", 13, alpha,
         word_count_actual / (real)(iter * train_words + 1) * 100,
         word_count_actual / ((real)(now - start + 1) / (real)CLOCKS_PER_SEC * 1000));
        fflush(stdout);
        if (word_count_actual / (real)(iter * train_words + 1) * 100 > 85) break;
      }
      alpha = starting_alpha * (1 - word_count_actual / (real)(iter * train_words + 1));
      if (alpha < starting_alpha * 0.0001) alpha = starting_alpha * 0.0001;
    }
    if (sentence_length == 0) {
      while (1) {
        word = ReadWordIndex(fi);
        if (feof(fi)) {
          break;
        }
        if (word == -1) {
          continue;
        }
        word_count++;
        if (sample > 0) {
          real ran = (sqrt(vocab[word].cn / (sample * train_words)) + 1) * (sample * train_words) / vocab[word].cn;
          next_random = next_random * (unsigned long long)25214903917 + 11;
          if (ran < (next_random & 0xFFFF) / (real)65536) continue;
        }
        sen[sentence_length] = word;
        sentence_length++;
        if (sentence_length >= MAX_SENTENCE_LENGTH) break;
      }
      sentence_position = 0;
    }
    if (feof(fi) || (word_count > train_words / num_threads)) {
      word_count_actual += word_count - last_word_count;
      local_iter--;
      if (local_iter == 0) break;
      word_count = 0;
      last_word_count = 0;
      sentence_length = 0;
      fseek(fi, file_size / (long long)num_threads * (long long)id, SEEK_SET);
      continue;
    }
    word = sen[sentence_position];
    if (word == -1) continue;
    for (c = 0; c < layer1_size; c++) neu1[c] = 0;
    next_random = next_random * (unsigned long long)25214903917 + 11;
    b = next_random % window;

    real cnt = 0;
    for (a = 0; a < layer1_size; ++a)
    	attention_syn[a] = 0;
    int window2 = 2;
    for (a = 0; a < window2 * 2 + 1; a++) if (a != window2) { // calc the context embedding(average of context word embeddings)
      c = sentence_position - window2 + a;
      if (c < 0) continue;
      if (c >= sentence_length) continue;
      last_word = sen[c];
      if (last_word == -1) continue;
      l1 = last_word * layer1_size;

      cnt += 1.0;
      for (c = 0; c < layer1_size; ++c)
      	attention_syn[c] += syn1neg[l1 + c];
  	}
  	if (cnt > 0) {
	  	for (a = 0; a < layer1_size; ++a)
	  		attention_syn[a] /= cnt;
      cnt = 1.0 / cnt;
  	}
    for (a = b; a < window * 2 + 1 - b; a++) if (a != window) {
      c = sentence_position - window + a;
      if (c < 0) continue;
      if (c >= sentence_length) continue;
      last_word = sen[c];
      if (last_word == -1) continue;
      l1 = last_word * layer1_size;

      // NEGATIVE SAMPLING
      if (negative > 0) for (d = 0; d < negative + 1; d++) {
        if (d == 0) {
          target = word;
          label = 1;
        } else {
          next_random = next_random * (unsigned long long)25214903917 + 11;
          target = table[(next_random >> 16) % table_size];
          if (target == 0) target = next_random % (vocab_size - 1) + 1;
          if (target == word) continue;
          label = 0;
        }
        l2 = target * layer1_size;
        f = 0;
        if (syn0[target].num == 1) {
          total = 1;
          for (p = 0; p < layer1_size; ++p) attention[p] = syn0[target].mult_sense_value[p];
        }
        else {
          real max_exp = -1000000000000;
          real min_exp = 1000000000000;
          real max_p = -1;
          for (p = 0; p < syn0[target].num; ++p) {
            // calc the average embedding of several sememes for each sense
            for (q = 0; q < layer1_size; ++q)
              avg_meaning[p * layer1_size + q] = 0;
            for (q = (p == 0 ? 0 : syn0[target].meaning_word_cnt[p - 1]); q < syn0[target].meaning_word_cnt[p]; ++q) {
              real *temp = &(meaning_syn[syn0[target].meaning_word_rank[q] * layer1_size]);
              for (c = 0; c < layer1_size; ++c) {
                avg_meaning[p * layer1_size + c] += temp[c];
              }
            }
            real divide_part;
            if (p == 0) {
              divide_part = 1.0 / (real)(syn0[target].meaning_word_cnt[0]);
            }
            else {
              divide_part = 1.0 / (real)(syn0[target].meaning_word_cnt[p] - syn0[target].meaning_word_cnt[p - 1]);
            }
            for (c = 0; c < layer1_size; ++c)
              avg_meaning[p * layer1_size + c] *= divide_part;
            // calc the weight for each sense
            _exp[p] = vectorDot(&(avg_meaning[p * layer1_size]), attention_syn, layer1_size);
            if (_exp[p] > max_exp) {
              max_exp = _exp[p]; 
              max_p = p;
            }
            if (_exp[p] < min_exp) {
              min_exp = _exp[p];
            }
          }
          real gap_exp = max_exp - min_exp;
          if (gap_exp >= 3.999) {
            gap_exp = 3.999;
          }
          total = pre_exp[(int)(gap_exp * 250.0)];
          // use the sense with max weight to represent a word
          for (p = 0; p < syn0[target].num; ++p) {
            if (p != max_p) {
              _exp[p] = 0.0;
              continue;
            }
            _exp[p] = 1.0;
            for (q = 0; q < layer1_size; ++q) {
              attention[q] = syn0[target].mult_sense_value[p * layer1_size + q];
            }
          }
        }
        // BP
        for (c = 0; c < layer1_size; c++) {
          f += syn1neg[c + l1] * attention[c];
        }
        if (f > MAX_EXP) 
          g = (label - 1) * alpha;
        else if (f < -MAX_EXP) 
          g = (label - 0) * alpha;
        else 
          g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha;
      	g /= total;
        for (c = 0; c < layer1_size; c++) {
          neu1e[c] = g * attention[c];
        }
        if (syn0[target].num == 1) {
          for (p = 0; p < layer1_size; ++p)
            syn0[target].mult_sense_value[p] += g * syn1neg[p + l1];
        }
        else {
          for (p = 0; p < layer1_size; ++p)
            mult_part[p] = g * syn1neg[p + l1] * attention_syn[p];
          real avg_num;
          for (p = 0; p < syn0[target].num; ++p) {
            if (_exp[p] < 0.1)
              continue;
            if (p == 0)
              avg_num = 1.0 / (real)(syn0[target].meaning_word_cnt[0]);
            else
              avg_num = 1.0 / (real)(syn0[target].meaning_word_cnt[p] - syn0[target].meaning_word_cnt[p - 1]);
            real *temp_mult = &(syn0[target].mult_sense_value[p * layer1_size]);
            for (c = 0; c < layer1_size; ++c) {
              mult_part2[c] = mult_part[c] * temp_mult[c] * avg_num;
            }
            for (q = (p == 0 ? 0 : syn0[target].meaning_word_cnt[p - 1]); q < syn0[target].meaning_word_cnt[p]; ++q) {
              real *temp = &(meaning_syn[syn0[target].meaning_word_rank[q] * layer1_size]);
              for (c = 0; c < layer1_size; ++c) {
                temp[c] += mult_part2[c];
              }
            }
            break;
          }

          for (c = 0; c < layer1_size; ++c)
              mult_part[c] = 0;
          for (p = 0; p < syn0[target].num; ++p) {
            real *temp_mult = &(syn0[target].mult_sense_value[p * layer1_size]);
            for (c = 0; c < layer1_size; ++c)
              mult_part[c] += avg_meaning[p * layer1_size + c] * temp_mult[c];
          }
          for (p = 0; p < layer1_size; ++p)
            mult_part[p] *= g * syn1neg[p + l1] * cnt;
          for (q = 0; q < window2 * 2 + 1; q++) if (q != window2) {
            p = sentence_position - window2 + q;
            if (p < 0) continue;
            if (p >= sentence_length) continue;
            p = sen[p];
            if (p == -1) continue;
            p = p * layer1_size;
            for (c = 0; c < layer1_size; ++c)
              syn1neg[p + c] += mult_part[c];
          }

          for (p = 0; p < syn0[target].num; ++p) {
            if (_exp[p] < 0.1)
              continue;
            real *temp_mult = &(syn0[target].mult_sense_value[p * layer1_size]);
            for (q = 0; q < layer1_size; ++q)
              temp_mult[q] += g * syn1neg[q + l1];
            break;
          }
        }

        for (c = 0; c < layer1_size; c++) {
          syn1neg[c + l1] += neu1e[c];
        }
      }
    }
    sentence_position++;
    if (sentence_position >= sentence_length) {
      sentence_length = 0;
      continue;
    }
  }

  fclose(fi);
  free(neu1);
  free(neu1e);

  free(attention);
  free(attention_syn);
  free(_exp);
  free(mult_part);

  printf("train end\n");

  pthread_exit(NULL);
}

void TrainModel() {
  long a, b, c, d;
  FILE *fo;
  pthread_t *pt = (pthread_t *)malloc(num_threads * sizeof(pthread_t));
  printf("Starting training using file %s\n", train_file);
  starting_alpha = alpha;
  if (read_vocab_file[0] != 0) ReadVocab();
  if (read_meaning_file[0] != 0) ReadMeaning();
  if (save_vocab_file[0] != 0) SaveVocab();
  if (output_file[0] == 0) return;
  InitNet();
  if (negative > 0) InitUnigramTable();
  if (checkpoint[0] != 0) ReadPoint();
  start = clock();
  for (a = 0; a < num_threads; a++) pthread_create(&pt[a], NULL, TrainModelThread, (void *)a);
  for (a = 0; a < num_threads; a++) pthread_join(pt[a], NULL);
  printf("all trains end\n");
  printf("%lld %lld %lld\n", meaning_size, vocab_size, layer1_size);
  fo = fopen(output_file, "w");
  if (classes == 0) {
    // Save the word vectors
    fprintf(fo, "%lld %lld %lld %f\n", meaning_size, vocab_size, layer1_size, alpha);
    for (a = 0; a < vocab_size; a++) {
      fprintf(fo, "%s ", vocab[a].word);
      fprintf(fo, "%d ", syn0[a].num);
      for (b = 0; b < syn0[a].num * layer1_size; b++) 
        fprintf(fo, "%lf ", syn0[a].mult_sense_value[b]);
      if (syn0[a].num > 1) {
        for (b = 0; b < syn0[a].num; ++b)
          fprintf(fo, "%d ", syn0[a].meaning_word_cnt[b]);
        for (b = 0; b < syn0[a].meaning_word_cnt[syn0[a].num - 1]; ++b)
          fprintf(fo, "%d ", syn0[a].meaning_word_rank[b]);
      }
      fprintf(fo, "\n");
    }
    for (a = 0; a < meaning_size; a++) {
      fprintf(fo, "%s ", meaning[a].word);
      for (b = 0; b < layer1_size; b++) 
        fprintf(fo, "%lf ", meaning_syn[a * layer1_size + b]);
      fprintf(fo, "\n");
    }
    for (a = 0; a < vocab_size * layer1_size; ++a)
      fprintf(fo, "%lf ", syn1neg[a]);
    fprintf(fo, "\n");
  }
  fclose(fo);
  printf("save end\n");
}

int ArgPos(char *str, int argc, char **argv) {
  int a;
  for (a = 1; a < argc; a++) if (!strcmp(str, argv[a])) {
    if (a == argc - 1) {
      printf("Argument missing for %s\n", str);
      exit(1);
    }
    return a;
  }
  return -1;
}

int main(int argc, char **argv) {
  int i;
  if (argc == 1) {
    printf("WORD VECTOR estimation toolkit v 0.1c\n\n");
    printf("Options:\n");
    printf("Parameters for training:\n");
    printf("\t-train <file>\n");
    printf("\t\tUse text data from <file> to train the model\n");
    printf("\t-output <file>\n");
    printf("\t\tUse <file> to save the resulting word vectors / word clusters\n");
    printf("\t-size <int>\n");
    printf("\t\tSet size of word vectors; default is 100\n");
    printf("\t-window <int>\n");
    printf("\t\tSet max skip length between words; default is 5\n");
    printf("\t-sample <float>\n");
    printf("\t\tSet threshold for occurrence of words. Those that appear with higher frequency in the training data\n");
    printf("\t\twill be randomly down-sampled; default is 1e-3, useful range is (0, 1e-5)\n");
    printf("\t-hs <int>\n");
    printf("\t\tUse Hierarchical Softmax; default is 0 (not used)\n");
    printf("\t-negative <int>\n");
    printf("\t\tNumber of negative examples; default is 5, common values are 3 - 10 (0 = not used)\n");
    printf("\t-threads <int>\n");
    printf("\t\tUse <int> threads (default 12)\n");
    printf("\t-iter <int>\n");
    printf("\t\tRun more training iterations (default 5)\n");
    printf("\t-min-count <int>\n");
    printf("\t\tThis will discard words that appear less than <int> times; default is 5\n");
    printf("\t-alpha <float>\n");
    printf("\t\tSet the starting learning rate; default is 0.025 for skip-gram and 0.05 for CBOW\n");
    printf("\t-classes <int>\n");
    printf("\t\tOutput word classes rather than word vectors; default number of classes is 0 (vectors are written)\n");
    printf("\t-debug <int>\n");
    printf("\t\tSet the debug mode (default = 2 = more info during training)\n");
    printf("\t-binary <int>\n");
    printf("\t\tSave the resulting vectors in binary moded; default is 0 (off)\n");
    printf("\t-save-vocab <file>\n");
    printf("\t\tThe vocabulary will be saved to <file>\n");
    printf("\t-read-vocab <file>\n");
    printf("\t\tThe vocabulary will be read from <file>, not constructed from the training data\n");
    printf("\t-cbow <int>\n");
    printf("\t\tUse the continuous bag of words model; default is 1 (use 0 for skip-gram model)\n");
    printf("\nExamples:\n");
    printf("./word2vec -train data.txt -output vec.txt -size 200 -window 5 -sample 1e-4 -negative 5 -hs 0 -binary 0 -cbow 1 -iter 3\n\n");
    return 0;
  }
  output_file[0] = 0;
  save_vocab_file[0] = 0;
  read_vocab_file[0] = 0;
  read_meaning_file[0] = 0;
  read_sense_file[0] = 0;
  checkpoint[0] = 0;
  if ((i = ArgPos((char *)"-size", argc, argv)) > 0) layer1_size = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-train", argc, argv)) > 0) strcpy(train_file, argv[i + 1]);
  if ((i = ArgPos((char *)"-save-vocab", argc, argv)) > 0) strcpy(save_vocab_file, argv[i + 1]);
  if ((i = ArgPos((char *)"-read-vocab", argc, argv)) > 0) strcpy(read_vocab_file, argv[i + 1]);
  if ((i = ArgPos((char *)"-read-meaning", argc, argv)) > 0) strcpy(read_meaning_file, argv[i + 1]);
  if ((i = ArgPos((char *)"-read-sense", argc, argv)) > 0) strcpy(read_sense_file, argv[i + 1]);
  if ((i = ArgPos((char *)"-checkpoint", argc, argv)) > 0) strcpy(checkpoint, argv[i + 1]);
  if ((i = ArgPos((char *)"-debug", argc, argv)) > 0) debug_mode = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-binary", argc, argv)) > 0) binary = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-cbow", argc, argv)) > 0) cbow = atoi(argv[i + 1]);
  if (cbow) alpha = 0.05;
  if ((i = ArgPos((char *)"-alpha", argc, argv)) > 0) alpha = atof(argv[i + 1]);
  if ((i = ArgPos((char *)"-output", argc, argv)) > 0) strcpy(output_file, argv[i + 1]);
  if ((i = ArgPos((char *)"-window", argc, argv)) > 0) window = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-sample", argc, argv)) > 0) sample = atof(argv[i + 1]);
  if ((i = ArgPos((char *)"-hs", argc, argv)) > 0) hs = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-negative", argc, argv)) > 0) negative = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-threads", argc, argv)) > 0) num_threads = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-iter", argc, argv)) > 0) iter = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-min-count", argc, argv)) > 0) min_count = atoi(argv[i + 1]);
  if ((i = ArgPos((char *)"-classes", argc, argv)) > 0) classes = atoi(argv[i + 1]);
  vocab = (struct vocab_word *)calloc(vocab_max_size, sizeof(struct vocab_word));
  meaning = (struct meaning_word *)calloc(2000, sizeof(struct meaning_word));
  vocab_hash = (int *)calloc(vocab_hash_size, sizeof(int));
  meaning_hash = (int *)calloc(meaning_hash_size, sizeof(int));
  expTable = (real *)malloc((EXP_TABLE_SIZE + 1) * sizeof(real));
  for (i = 0; i < EXP_TABLE_SIZE; i++) {
    expTable[i] = exp((i / (real)EXP_TABLE_SIZE * 2 - 1) * MAX_EXP); // Precompute the exp() table
    expTable[i] = expTable[i] / (expTable[i] + 1);                   // Precompute f(x) = x / (x + 1)
  }
  for (i = 0; i < EXP_FROM_ZERO_FORE; ++i)
    pre_exp[i] = exp((real)i / (real)(EXP_FROM_ZERO_FORE / 4));
  TrainModel();
  return 0;
}