Transformer.cpp/tokenizer.cpp at main · projektjoe/Transformer.cpp · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
#include "tokenizer.h"
#include <iostream>
#include <fstream>
#include <algorithm>
#include <cstdio>
#include <cctype>

bool Tokenizer::compareTokens(const TokenIndex& a, const TokenIndex& b) {
    return a.str < b.str;
}

Tokenizer::Tokenizer(const std::string& tokenizer_path, int vocab_size)
        : vocab_size(vocab_size) {
    vocab.resize(vocab_size);
    vocab_scores.resize(vocab_size);

    for (int i = 0; i < 256; i++) {
        byte_pieces[i * 2] = static_cast<char>(i);
        byte_pieces[i * 2 + 1] = '\0';
    }

    std::ifstream file(tokenizer_path, std::ios::binary);
    if (!file) {
        std::cerr << "Couldn't load " << tokenizer_path << std::endl;
        exit(EXIT_FAILURE);
    }

    file.read(reinterpret_cast<char*>(&max_token_length), sizeof(int));

    for (int i = 0; i < vocab_size; i++) {
        file.read(reinterpret_cast<char*>(&vocab_scores[i]), sizeof(float));
        int len;
        file.read(reinterpret_cast<char*>(&len), sizeof(int));
        vocab[i].resize(len);
        file.read(&vocab[i][0], len);
    }

    buildSortedVocab();
}

void Tokenizer::buildSortedVocab() {
    sorted_vocab.resize(vocab_size);
    for (int i = 0; i < vocab_size; i++) {
        sorted_vocab[i] = {vocab[i], i};
    }
    std::sort(sorted_vocab.begin(), sorted_vocab.end(), compareTokens);
}

std::string Tokenizer::decode(int prev_token, int token) {
    std::string piece = vocab[token];
    if (prev_token == 1 && piece[0] == ' ') {
        piece = piece.substr(1);
    }

    unsigned char byte_val;
    if (sscanf(piece.c_str(), "<0x%2hhX>", &byte_val) == 1) {
        piece = std::string(byte_pieces.data() + byte_val * 2, 1);
    }

    return piece;
}

void Tokenizer::safePrintf(const std::string& piece) {
    if (piece.empty()) return;
    if (piece.length() == 1) {
        unsigned char byte_val = piece[0];
        if (!(isprint(byte_val) || isspace(byte_val))) {
            return;
        }
    }
    std::cout << piece;
}

int Tokenizer::strLookup(const std::string& str) {
    TokenIndex tok{str, 0};
    auto it = std::lower_bound(sorted_vocab.begin(), sorted_vocab.end(), tok,
                               [](const TokenIndex& a, const TokenIndex& b) {
                                   return a.str < b.str;
                               });
    if (it != sorted_vocab.end() && it->str == str) {
        return it->id;
    }
    return -1;
}

std::vector<int> Tokenizer::encode(const std::string& text, bool bos, bool eos) {
    std::vector<int> tokens;
    tokens.reserve(text.length() + 2);  // +2 for potential BOS and EOS tokens

    if (bos) tokens.push_back(1);

    if (!text.empty()) {
        int dummy_prefix = strLookup(" ");
        tokens.push_back(dummy_prefix);
    }

    std::string str_buffer;
    str_buffer.reserve(max_token_length * 2 + 1 + 2);

    for (size_t i = 0; i < text.length(); ++i) {
        if ((text[i] & 0xC0) != 0x80) {
            str_buffer.clear();
        }

        str_buffer += text[i];

        if (i + 1 == text.length() || (text[i + 1] & 0xC0) != 0x80 || str_buffer.length() >= 4) {
            int id = strLookup(str_buffer);
            if (id != -1) {
                tokens.push_back(id);
            } else {
                for (char c : str_buffer) {
                    tokens.push_back(static_cast<unsigned char>(c) + 3);
                }
            }
            str_buffer.clear();
        }
    }

    while (true) {
        float best_score = -1e10;
        int best_id = -1;
        int best_idx = -1;

        for (size_t i = 0; i < tokens.size() - 1; ++i) {
            std::string merge_str = vocab[tokens[i]] + vocab[tokens[i + 1]];
            int id = strLookup(merge_str);
            if (id != -1 && vocab_scores[id] > best_score) {
                best_score = vocab_scores[id];
                best_id = id;
                best_idx = i;
            }
        }

        if (best_idx == -1) break;

        tokens[best_idx] = best_id;
        tokens.erase(tokens.begin() + best_idx + 1);
    }

    if (eos) tokens.push_back(2);

    return tokens;
}