llama : fix BPE pre-tokenization (#6920)

* merged the changes from deepseeker models to main branch

* Moved regex patterns to unicode.cpp and updated unicode.h

* Moved header files

* Resolved issues

* added and refactored unicode_regex_split and related functions

* Updated/merged the deepseek coder pr

* Refactored code

* Adding unicode regex mappings

* Adding unicode regex function

* Added needed functionality, testing remains

* Fixed issues

* Fixed issue with gpt2 regex custom preprocessor

* unicode : fix? unicode_wstring_to_utf8

* lint : fix whitespaces

* tests : add tokenizer tests for numbers

* unicode : remove redundant headers

* tests : remove and rename tokenizer test scripts

* tests : add sample usage

* gguf-py : reader prints warnings on duplicate keys

* llama : towards llama3 tokenization support (wip)

* unicode : shot in the dark to fix tests on Windows

* unicode : first try custom implementations

* convert : add "tokenizer.ggml.pre" GGUF KV (wip)

* llama : use new pre-tokenizer type

* convert : fix pre-tokenizer type writing

* lint : fix

* make : add test-tokenizer-0-llama-v3

* wip

* models : add llama v3 vocab file

* llama : adapt punctuation regex + add llama 3 regex

* minor

* unicode : set bomb

* unicode : set bomb

* unicode : always use std::wregex

* unicode : support \p{N}, \p{L} and \p{P} natively

* unicode : try fix windows

* unicode : category support via std::regex

* unicode : clean-up

* unicode : simplify

* convert : add convert-hf-to-gguf-update.py

ggml-ci

* lint : update

* convert : add falcon

ggml-ci

* unicode : normalize signatures

* lint : fix

* lint : fix

* convert : remove unused functions

* convert : add comments

* convert : exercise contractions

ggml-ci

* lint : fix

* cmake : refactor test targets

* tests : refactor vocab tests

ggml-ci

* tests : add more vocabs and tests

ggml-ci

* unicode : cleanup

* scripts : ignore new update script in check-requirements.sh

* models : add phi-3, mpt, gpt-2, starcoder

* tests : disable obsolete

ggml-ci

* tests : use faster bpe test

ggml-ci

* llama : more prominent warning for old BPE models

* tests : disable test-tokenizer-1-bpe due to slowness

ggml-ci

---------

Co-authored-by: Jaggzh <jaggz.h@gmail.com>
Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>

1 files changed, 448 insertions, 39 deletions

diff --git a/unicode.cpp b/unicode.cpp
index df8c5f58..f2ccda05 100644
--- a/unicode.cpp
+++ b/unicode.cpp
@@ -5,11 +5,14 @@
 #include <cstddef>
 #include <cstdint>
 #include <map>
+#include <regex>
 #include <stdexcept>
 #include <string>
 #include <unordered_map>
 #include <utility>
 #include <vector>
+#include <locale>
+#include <codecvt>
 
 static std::string unicode_cpts_to_utf8(const std::vector<uint32_t> & cps) {
     std::string result;
@@ -53,23 +56,22 @@ static uint32_t unicode_cpt_from_utf8(const std::string & utf8, size_t & offset)
         offset += 4;
         return result;
     }
-    throw std::invalid_argument("invalid string");
+    throw std::invalid_argument("failed to convert utf8 to codepoint");
 }
 
-static std::vector<uint16_t> unicode_cpt_to_utf16(uint32_t cp) {
-    std::vector<uint16_t> result;
-    if (/* 0x0000 <= cp && */ cp <= 0xffff) {
-        result.emplace_back(cp);
-    }
-    else if (0x10000 <= cp && cp <= 0x10ffff) {
-        result.emplace_back(0xd800 | ((cp - 0x10000) >> 10));
-        result.emplace_back(0xdc00 | ((cp - 0x10000) & 0x03ff));
-    }
-    else {
-        throw std::invalid_argument("invalid cpt");
-    }
-    return result;
-}
+//static std::vector<uint16_t> unicode_cpt_to_utf16(uint32_t cp) {
+//    std::vector<uint16_t> result;
+//    if (/* 0x0000 <= cp && */ cp <= 0xffff) {
+//        result.emplace_back(cp);
+//        return result;
+//    }
+//    if (0x10000 <= cp && cp <= 0x10ffff) {
+//        result.emplace_back(0xd800 | ((cp - 0x10000) >> 10));
+//        result.emplace_back(0xdc00 | ((cp - 0x10000) & 0x03ff));
+//        return result;
+//    }
+//    throw std::invalid_argument("failed to convert codepoint to utf16");
+//}
 
 //static std::vector<uint16_t> unicode_cpts_to_utf16(const std::vector<uint32_t> & cps) {
 //    std::vector<uint16_t> result;
@@ -80,28 +82,28 @@ static std::vector<uint16_t> unicode_cpt_to_utf16(uint32_t cp) {
 //    return result;
 //}
 
-static uint32_t cpt_from_utf16(const std::vector<uint16_t> & utf16, size_t & offset) {
-    assert(offset < utf16.size());
-    if (((utf16[0] >> 10) << 10) != 0xd800) {
-        auto result = utf16[offset + 0];
-        offset += 1;
-        return result;
-    }
-
-    if (offset + 1 >= utf16.size() || !((utf16[1] & 0xdc00) == 0xdc00)) {
-        throw std::invalid_argument("invalid character");
-    }
-
-    auto result = 0x10000 + (((utf16[0] & 0x03ff) << 10) | (utf16[1] & 0x03ff));
-    offset += 2;
-    return result;
-}
+//static uint32_t unicode_cpt_from_utf16(const std::vector<uint16_t> & utf16, size_t & offset) {
+//    assert(offset < utf16.size());
+//    if (((utf16[0] >> 10) << 10) != 0xd800) {
+//        auto result = utf16[offset + 0];
+//        offset += 1;
+//        return result;
+//    }
+//
+//    if (offset + 1 >= utf16.size() || !((utf16[1] & 0xdc00) == 0xdc00)) {
+//        throw std::invalid_argument("invalid character");
+//    }
+//
+//    auto result = 0x10000 + (((utf16[0] & 0x03ff) << 10) | (utf16[1] & 0x03ff));
+//    offset += 2;
+//    return result;
+//}
 
 //static std::vector<uint32_t> unicode_cpts_from_utf16(const std::vector<uint16_t> & utf16) {
 //    std::vector<uint32_t> result;
 //    size_t offset = 0;
 //    while (offset < utf16.size()) {
-//        result.push_back(cpt_from_utf16(utf16, offset));
+//        result.push_back(unicode_cpt_from_utf16(utf16, offset));
 //    }
 //    return result;
 //}
@@ -194,34 +196,277 @@ static std::unordered_map<std::string, uint8_t> unicode_utf8_to_byte_map() {
     return map;
 }
 
+static inline std::wstring unicode_wstring_from_utf8(const std::string & s) {
+    std::wstring_convert<std::codecvt_utf8<wchar_t>> conv;
+    return conv.from_bytes(s);
+}
+
+static std::vector<std::string> unicode_byte_encoding_process(const std::vector<std::string> & bpe_words) {
+    std::vector<std::string> bpe_encoded_words;
+    for (const auto & word : bpe_words) {
+        std::string text_utf;
+        auto utf_word =  unicode_cpts_from_utf8(word);
+        for (size_t i = 0; i < utf_word.size(); ++i) {
+            text_utf += unicode_cpt_to_utf8(utf_word[i]);
+        }
+
+        std::string encoded_token;
+        for (char & c : text_utf) {
+            encoded_token += unicode_byte_to_utf8(c);
+        }
+        bpe_encoded_words.emplace_back(encoded_token);
+    }
+    return bpe_encoded_words;
+}
+
+// GPT2 system regex:  's|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+
+static std::vector<size_t> unicode_regex_split_custom_gpt2(const std::string & text, const std::vector<size_t> & offsets) {
+    std::vector<size_t> bpe_offsets; // store the offset of each word
+    bpe_offsets.reserve(offsets.size()); // Reserve memory for the approximate size
+
+    size_t start = 0;
+
+    const auto cpts = unicode_cpts_from_utf8(text);
+
+    for (auto offset : offsets) {
+        std::string token;
+
+        bool collecting_numeric = false;
+        bool collecting_letter = false;
+        bool collecting_special = false;
+        bool collecting_whitespace_lookahead = false;
+        bool collecting = false;
+
+        std::vector<std::string> text_utf;
+        text_utf.reserve(offset);
+
+        for (size_t i = start; i < start + offset; ++i) {
+            text_utf.emplace_back(unicode_cpt_to_utf8(cpts[i]));
+        }
+
+        for (int i = 0; i < (int)text_utf.size(); i++) {
+            const std::string & utf_char = text_utf[i];
+            bool split_condition = false;
+            int bytes_remain = text_utf.size() - i;
+
+            // forward backward lookups
+            const std::string & utf_char_next      = (i + 1 < (int)text_utf.size()) ? text_utf[i + 1] : "";
+            const std::string & utf_char_next_next = (i + 2 < (int)text_utf.size()) ? text_utf[i + 2] : "";
+
+            // handling contractions
+            if (!split_condition && bytes_remain >= 2) {
+                // 's|'t|'m|'d
+                if (utf_char == "\'" && (utf_char_next == "s" || utf_char_next == "t" || utf_char_next == "m" || utf_char_next == "d")) {
+                    split_condition = true;
+                }
+                if (split_condition) {
+                    if (token.size()) {
+                        bpe_offsets.emplace_back(unicode_cpts_from_utf8(token).size());
+                    }
+                    token = utf_char + utf_char_next;
+                    bpe_offsets.emplace_back(unicode_cpts_from_utf8(token).size());
+                    token = "";
+                    i++;
+                    continue;
+                }
+            }
+            if (!split_condition && bytes_remain >= 3) {
+                // 're|'ve|'ll
+                if (utf_char == "\'" && (
+                    (utf_char_next == "r" && utf_char_next_next == "e") ||
+                    (utf_char_next == "v" && utf_char_next_next == "e") ||
+                    (utf_char_next == "l" && utf_char_next_next == "l"))
+                    ) {
+                    split_condition = true;
+                }
+                if (split_condition) {
+                    // current token + next token can be defined
+                    if (token.size()) {
+                        bpe_offsets.emplace_back(unicode_cpts_from_utf8(token).size());
+                    }
+                    token =  utf_char;
+                    token += utf_char_next;
+                    token += utf_char_next_next;
+
+                    bpe_offsets.emplace_back(unicode_cpts_from_utf8(token).size());
+                    token = "";
+                    i += 2;
+                    continue;
+                }
+            }
+
+            if (!split_condition && !collecting) {
+                if (unicode_cpt_type(utf_char) == CODEPOINT_TYPE_LETTER || (token.empty() && utf_char == " " && unicode_cpt_type(utf_char_next) == CODEPOINT_TYPE_LETTER)) {
+                    collecting_letter = true;
+                    collecting = true;
+                }
+                else if (unicode_cpt_type(utf_char) == CODEPOINT_TYPE_DIGIT || (token.empty() && utf_char == " " && unicode_cpt_type(utf_char_next) == CODEPOINT_TYPE_DIGIT)) {
+                    collecting_numeric = true;
+                    collecting = true;
+                }
+                else if (
+                    ((unicode_cpt_type(utf_char) != CODEPOINT_TYPE_LETTER && unicode_cpt_type(utf_char) != CODEPOINT_TYPE_DIGIT) && (unicode_cpt_type(utf_char) != CODEPOINT_TYPE_WHITESPACE)) ||
+                    (token.empty() && utf_char == " " && unicode_cpt_type(utf_char_next) != CODEPOINT_TYPE_LETTER && unicode_cpt_type(utf_char_next) != CODEPOINT_TYPE_DIGIT && unicode_cpt_type(utf_char_next) != CODEPOINT_TYPE_WHITESPACE)
+                    ) {
+                    collecting_special = true;
+                    collecting = true;
+                }
+                else if (unicode_cpt_type(utf_char) == CODEPOINT_TYPE_WHITESPACE && unicode_cpt_type(utf_char_next) == CODEPOINT_TYPE_WHITESPACE) {
+                    collecting_whitespace_lookahead = true;
+                    collecting = true;
+                }
+                else if (unicode_cpt_type(utf_char) == CODEPOINT_TYPE_WHITESPACE) {
+                    split_condition = true;
+                }
+            }
+            else if (!split_condition && collecting) {
+                if (collecting_letter && unicode_cpt_type(utf_char) != CODEPOINT_TYPE_LETTER) {
+                    split_condition = true;
+                }
+                else if (collecting_numeric && unicode_cpt_type(utf_char) != CODEPOINT_TYPE_DIGIT) {
+                    split_condition = true;
+                }
+                else if (collecting_special && (unicode_cpt_type(utf_char) == CODEPOINT_TYPE_LETTER || unicode_cpt_type(utf_char) == CODEPOINT_TYPE_DIGIT || unicode_cpt_type(utf_char) == CODEPOINT_TYPE_WHITESPACE)) {
+                    split_condition = true;
+                }
+                else if (collecting_whitespace_lookahead && (unicode_cpt_type(utf_char_next) == CODEPOINT_TYPE_LETTER || unicode_cpt_type(utf_char_next) == CODEPOINT_TYPE_DIGIT)) {
+                    split_condition = true;
+                }
+            }
+
+            if (utf_char_next == "") {
+                split_condition = true; // final
+                token += utf_char;
+            }
+
+            if (split_condition) {
+                if (token.size()) {
+                    bpe_offsets.emplace_back(unicode_cpts_from_utf8(token).size());
+                }
+                token = utf_char;
+                collecting = false;
+                collecting_letter = false;
+                collecting_numeric = false;
+                collecting_special = false;
+                collecting_whitespace_lookahead = false;
+            }
+            else {
+                token += utf_char;
+            }
+        }
+
+        start += offset;
+    }
+
+    return bpe_offsets;
+}
+
+// use std::wregex to split the text
+static std::vector<size_t> unicode_regex_split_stl(const std::wstring & wtext, const std::wstring & regex_expr, const std::vector<size_t> & offsets) {
+    std::wregex expr(regex_expr);
+    std::vector<size_t> bpe_offsets; // store the offset of each word
+    bpe_offsets.reserve(offsets.size()); // Reserve memory for the approximate size
+    size_t start = 0;
+    for (auto offset : offsets) {
+        std::wcregex_iterator it(wtext.data() + start, wtext.data() + start + offset, expr);
+        std::wcregex_iterator end;
+
+        int64_t start_idx = 0;
+        while (it != end) {
+            std::wcmatch match = *it;
+            if (match.position() > start_idx) {
+                bpe_offsets.emplace_back(match.position() - start_idx);
+            }
+            bpe_offsets.emplace_back(match.length());
+            start_idx = match.position() + match.length();
+            ++it;
+        }
+
+        if (start_idx < (int64_t) offset) {
+            bpe_offsets.emplace_back(offset - start_idx);
+        }
+        start += offset;
+    }
+
+    return bpe_offsets;
+}
+
+// use std::regex to split the text
+static std::vector<size_t> unicode_regex_split_stl(const std::string & text, const std::string & regex_expr, const std::vector<size_t> & offsets) {
+    std::regex expr(regex_expr);
+    std::vector<size_t> bpe_offsets; // store the offset of each word
+    bpe_offsets.reserve(offsets.size()); // Reserve memory for the approximate size
+    size_t start = 0;
+    for (auto offset : offsets) {
+        std::cregex_iterator it(text.data() + start, text.data() + start + offset, expr);
+        std::cregex_iterator end;
+
+        int64_t start_idx = 0;
+        while (it != end) {
+            std::cmatch match = *it;
+            if (match.position() > start_idx) {
+                bpe_offsets.emplace_back(match.position() - start_idx);
+            }
+            bpe_offsets.emplace_back(match.length());
+            start_idx = match.position() + match.length();
+            ++it;
+        }
+
+        if (start_idx < (int64_t) offset) {
+            bpe_offsets.emplace_back(offset - start_idx);
+        }
+        start += offset;
+    }
+
+    return bpe_offsets;
+}
+
+static std::vector<size_t> unicode_regex_split_custom(const std::string & text, const std::string & regex_expr, const std::vector<size_t> & offsets) {
+    std::vector<size_t> bpe_offsets;
+
+    (void)(text);
+    (void)(regex_expr);
+    (void)(offsets);
+    // TODO: this implementation is actually wrong, uncomment and run:
+    //       make -j && ./bin/test-tokenizer-0 ../models/ggml-vocab-gpt-2.gguf
+    //if (regex_expr == "'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)") {
+    //    bpe_offsets = unicode_regex_split_custom_gpt2(text, offsets);
+    //}
+
+    return bpe_offsets;
+}
+
 //
 // interface
 //
 
 std::string unicode_cpt_to_utf8(uint32_t cp) {
     std::string result;
+
     if (/* 0x00 <= cp && */ cp <= 0x7f) {
         result.push_back(cp);
+        return result;
     }
-    else if (0x80 <= cp && cp <= 0x7ff) {
+    if (0x80 <= cp && cp <= 0x7ff) {
         result.push_back(0xc0 | ((cp >> 6) & 0x1f));
         result.push_back(0x80 | (cp & 0x3f));
+        return result;
     }
-    else if (0x800 <= cp && cp <= 0xffff) {
+    if (0x800 <= cp && cp <= 0xffff) {
         result.push_back(0xe0 | ((cp >> 12) & 0x0f));
         result.push_back(0x80 | ((cp >> 6) & 0x3f));
         result.push_back(0x80 | (cp & 0x3f));
+        return result;
     }
-    else if (0x10000 <= cp && cp <= 0x10ffff) {
+    if (0x10000 <= cp && cp <= 0x10ffff) {
         result.push_back(0xf0 | ((cp >> 18) & 0x07));
         result.push_back(0x80 | ((cp >> 12) & 0x3f));
         result.push_back(0x80 | ((cp >> 6) & 0x3f));
         result.push_back(0x80 | (cp & 0x3f));
+        return result;
     }
-    else {
-        throw std::invalid_argument("invalid codepoint");
-    }
-    return result;
+
+    throw std::invalid_argument("invalid codepoint");
 }
 
 std::vector<uint32_t> unicode_cpts_normalize_nfd(const std::vector<uint32_t> & cpts) {
@@ -275,3 +520,167 @@ char32_t unicode_tolower(char32_t cp) {
     auto it = unicode_map_lowercase.find(cp);
     return it == unicode_map_lowercase.end() ? cp : it->second;
 }
+
+std::vector<std::string> unicode_regex_split(const std::string & text, const std::vector<std::string> & regex_exprs) {
+    // unicode categories
+    static const std::map<std::string, int> k_ucat_enum = {
+        { "\\p{N}", CODEPOINT_TYPE_DIGIT },
+        { "\\p{L}", CODEPOINT_TYPE_LETTER },
+        { "\\p{P}", CODEPOINT_TYPE_PUNCTUATION },
+    };
+
+    static const std::map<int, int> k_ucat_cpt = {
+        { CODEPOINT_TYPE_DIGIT,         0xD1 },
+        { CODEPOINT_TYPE_LETTER,        0xD2 },
+        { CODEPOINT_TYPE_PUNCTUATION,   0xD3 },
+    };
+
+    static const std::map<int, std::string> k_ucat_map = {
+        { CODEPOINT_TYPE_DIGIT,         "\x30-\x39" }, // 0-9
+        { CODEPOINT_TYPE_LETTER,        "\x41-\x5A\x61-\x7A" }, // A-Za-z
+        { CODEPOINT_TYPE_PUNCTUATION,   "\x21-\x23\x25-\x2A\x2C-\x2F\x3A-\x3B\x3F-\x40\\\x5B-\\\x5D\x5F\\\x7B\\\x7D" }, // !-#%-*,-/:-;?-@\[-\]_\{\}
+    };
+
+    // compute collapsed codepoints only if needed by at least one regex
+    bool need_collapse = false;
+    for (auto & regex_expr : regex_exprs) {
+        // search for unicode categories
+        for (const auto & ucat : k_ucat_enum) {
+            if (std::string::npos != regex_expr.find(ucat.first)) {
+                need_collapse = true;
+                break;
+            }
+        }
+    }
+
+    const auto cpts = unicode_cpts_from_utf8(text);
+
+    // generate a "collapsed" representation of the text, where all codepoints are replaced by a single byte
+    // ref: https://github.com/ggerganov/llama.cpp/pull/6920#issuecomment-2081479935
+    std::string text_collapsed;
+    if (need_collapse) {
+        // collapse all unicode categories
+        text_collapsed.resize(cpts.size());
+
+        for (size_t i = 0; i < cpts.size(); ++i) {
+            // keep single-byte codepoints as is
+            if (cpts[i] < 128) {
+                text_collapsed[i] = cpts[i];
+                continue;
+            }
+
+            const int cpt_type = unicode_cpt_type(cpts[i]);
+
+            if (k_ucat_cpt.find(cpt_type) != k_ucat_cpt.end()) {
+                text_collapsed[i] = k_ucat_cpt.at(cpt_type);
+            } else {
+                text_collapsed[i] = (char) 0xD0; // fallback
+            }
+        }
+    }
+
+    std::vector<size_t> bpe_offsets = { cpts.size() };
+
+    for (auto & regex_expr : regex_exprs) {
+        // first, see if we have an efficient custom regex implementation
+        auto tmp = unicode_regex_split_custom(text, regex_expr, bpe_offsets);
+
+        if (!tmp.empty()) {
+            bpe_offsets = std::move(tmp);
+            continue;
+        }
+
+        // fallback to general-purpose std::regex / std::wregex
+        try {
+            // if a unicode category is used in the regex, we use the collapsed text and replace the unicode category
+            // with the corresponding collapsed representation
+            bool use_collapsed = false;
+            for (auto & ucat : k_ucat_enum) {
+                if (std::string::npos != regex_expr.find(ucat.first)) {
+                    use_collapsed = true;
+                    break;
+                }
+            }
+
+            if (use_collapsed) {
+                // sanity-check that the original regex does not contain any non-ASCII characters
+                const auto cpts_regex = unicode_cpts_from_utf8(regex_expr);
+                for (size_t i = 0; i < cpts_regex.size(); ++i) {
+                    if (cpts_regex[i] >= 128) {
+                        throw std::runtime_error("Regex includes both unicode categories and non-ASCII characters - not supported");
+                    }
+                }
+
+                // generate a collapsed representation of the regex
+                std::string regex_expr_collapsed;
+
+                // track if we are inside [], because nested [] are not allowed
+                bool inside = false;
+                for (size_t i = 0; i < regex_expr.size(); ++i) {
+                    if (regex_expr[i] == '[' && (i == 0 || regex_expr[i - 1] != '\\')) {
+                        regex_expr_collapsed += '[';
+                        inside = true;
+                        continue;
+                    }
+
+                    if (inside && regex_expr[i] == ']' && regex_expr[i - 1] != '\\') {
+                        regex_expr_collapsed += ']';
+                        inside = false;
+                        continue;
+                    }
+
+                    if (regex_expr[i + 0] == '\\' && i + 4 < regex_expr.size() &&
+                        regex_expr[i + 1] == 'p' &&
+                        regex_expr[i + 2] == '{' &&
+                        regex_expr[i + 4] == '}') {
+                        const std::string pat = regex_expr.substr(i, 5);
+                        if (k_ucat_enum.find(pat) != k_ucat_enum.end()) {
+                            if (!inside) {
+                                regex_expr_collapsed += '[';
+                            }
+                            regex_expr_collapsed += k_ucat_cpt.at(k_ucat_enum.at(pat));
+                            regex_expr_collapsed += k_ucat_map.at(k_ucat_enum.at(pat));
+                            if (!inside) {
+                                regex_expr_collapsed += ']';
+                            }
+                            i += 4;
+                            continue;
+                        }
+                    }
+
+                    regex_expr_collapsed += regex_expr[i];
+                }
+
+                //printf("text_collapsed: %s\n", text_collapsed.c_str());
+                //printf("regex_expr_collapsed: %s\n", regex_expr_collapsed.c_str());
+                bpe_offsets = unicode_regex_split_stl(text_collapsed, regex_expr_collapsed, bpe_offsets);
+            } else {
+                // no unicode category used, we can use std::wregex directly
+                const std::wstring wtext       = unicode_wstring_from_utf8(text);
+                const std::wstring wregex_expr = unicode_wstring_from_utf8(regex_expr);
+
+                //printf("text: %s\n", text.c_str());
+                //printf("regex_expr: %s\n", regex_expr.c_str());
+                bpe_offsets = unicode_regex_split_stl(wtext, wregex_expr, bpe_offsets);
+            }
+        } catch (std::regex_error & e) {
+            fprintf(stderr, "Failed to process regex: '%s'\n", regex_expr.c_str());
+            fprintf(stderr, "Regex error: %s\n", e.what());
+            throw std::runtime_error("Failed to process regex");
+        }
+    }
+
+    std::vector<std::string> bpe_words;
+    bpe_words.reserve(bpe_offsets.size()); // reserve memory for the approximate size
+
+    size_t start = 0;
+    for (size_t & offset : bpe_offsets) {
+        bpe_words.emplace_back();
+        for (size_t i = start; i < start + offset; ++i) {
+            bpe_words.back() += unicode_cpt_to_utf8(cpts[i]);
+        }
+        start += offset;
+    }
+
+    return unicode_byte_encoding_process(bpe_words);
+}