fix: enhance the tokenizer's handing of Unicode (leejet#120)

Author: leejet

model.cpp

@@ -1192,20 +1192,9 @@ ggml_type ModelLoader::get_sd_wtype() {
     return GGML_TYPE_COUNT;
 }
 
-bool ModelLoader::load_vocab(on_new_token_cb_t on_new_token_cb) {
-    char* vocab_buffer          = reinterpret_cast<char*>(vocab_json);
-    nlohmann::json vocab        = nlohmann::json::parse(vocab_buffer);
-    std::map<char, int> decoder = unicode_to_byte();
-    for (auto& it : vocab.items()) {
-        int token_id          = it.value();
-        std::string token_str = it.key();
-        std::string token     = "";
-        for (char c : token_str) {
-            token += decoder[c];
-        }
-        on_new_token_cb(token, token_id);
-    }
-    return true;
+std::string ModelLoader::load_merges() {
+    std::string merges_utf8_str(reinterpret_cast<const char*>(merges_utf8_c_str), sizeof(merges_utf8_c_str));
+    return merges_utf8_str;
 }
 
 bool ModelLoader::load_tensors(on_new_tensor_cb_t on_new_tensor_cb) {

model.h

@@ -115,7 +115,7 @@ class ModelLoader {
     bool init_from_file(const std::string& file_path, const std::string& prefix = "");
     SDVersion get_sd_version();
     ggml_type get_sd_wtype();
-    bool load_vocab(on_new_token_cb_t on_new_token_cb);
+    std::string load_merges();
     bool load_tensors(on_new_tensor_cb_t on_new_tensor_cb);
     int64_t cal_mem_size(ggml_backend_t backend);
     ~ModelLoader() = default;

stable-diffusion.cpp

@@ -493,12 +493,40 @@ const int BOS_TOKEN_ID = 49406;
 const int EOS_TOKEN_ID = 49407;
 const int PAD_TOKEN_ID = 49407;
 
+std::vector<std::pair<int, std::u32string>> bytes_to_unicode() {
+    std::vector<std::pair<int, std::u32string>> byte_unicode_pairs;
+    std::set<int> byte_set;
+    for (int b = static_cast<int>('!'); b <= static_cast<int>('~'); ++b) {
+        byte_set.insert(b);
+        byte_unicode_pairs.push_back(std::pair<int, std::u32string>(b, unicode_value_to_utf32(b)));
+    }
+    for (int b = 161; b <= 172; ++b) {
+        byte_set.insert(b);
+        byte_unicode_pairs.push_back(std::pair<int, std::u32string>(b, unicode_value_to_utf32(b)));
+    }
+    for (int b = 174; b <= 255; ++b) {
+        byte_set.insert(b);
+        byte_unicode_pairs.push_back(std::pair<int, std::u32string>(b, unicode_value_to_utf32(b)));
+    }
+    int n = 0;
+    for (int b = 0; b < 256; ++b) {
+        if (byte_set.find(b) == byte_set.end()) {
+            byte_unicode_pairs.push_back(std::pair<int, std::u32string>(b, unicode_value_to_utf32(n + 256)));
+            ++n;
+        }
+    }
+    // LOG_DEBUG("byte_unicode_pairs %d", byte_unicode_pairs.size());
+    return byte_unicode_pairs;
+}
+
 // Ref: https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
 // TODO: implement bpe
 class CLIPTokenizer {
 private:
     SDVersion version = VERSION_1_x;
-    std::map<std::string, int32_t> encoder;
+    std::map<int, std::u32string> byte_encoder;
+    std::map<std::u32string, int> encoder;
+    std::map<std::pair<std::u32string, std::u32string>, int> bpe_ranks;
     std::regex pat;
 
     static std::string strip(const std::string& str) {
@@ -521,19 +549,61 @@ class CLIPTokenizer {
 
 public:
     CLIPTokenizer(SDVersion version = VERSION_1_x)
-        : version(version){};
-    std::string bpe(std::string token) {
-        std::string word = token + "</w>";
+        : version(version) {}
+
+    void load_from_merges(const std::string& merges_utf8_str) {
+        auto byte_unicode_pairs = bytes_to_unicode();
+        byte_encoder            = std::map<int, std::u32string>(byte_unicode_pairs.begin(), byte_unicode_pairs.end());
+        // for (auto & pair: byte_unicode_pairs) {
+        //     std::cout << pair.first << ": " << pair.second << std::endl;
+        // }
+        std::vector<std::u32string> merges;
+        size_t start = 0;
+        size_t pos;
+        std::u32string merges_utf32_str = utf8_to_utf32(merges_utf8_str);
+        while ((pos = merges_utf32_str.find('\n', start)) != std::string::npos) {
+            merges.push_back(merges_utf32_str.substr(start, pos - start));
+            start = pos + 1;
+        }
+        merges = std::vector<std::u32string>(merges.begin() + 1, merges.begin() + 49152 - 256 - 2 + 1);
+        std::vector<std::pair<std::u32string, std::u32string>> merge_pairs;
+        for (const auto& merge : merges) {
+            size_t space_pos = merge.find(' ');
+            merge_pairs.emplace_back(merge.substr(0, space_pos), merge.substr(space_pos + 1));
+            // LOG_DEBUG("%s", utf32_to_utf8(merge.substr(space_pos + 1)).c_str());
+        }
+        std::vector<std::u32string> vocab;
+        for (const auto& pair : byte_unicode_pairs) {
+            vocab.push_back(pair.second);
+        }
+        for (const auto& pair : byte_unicode_pairs) {
+            vocab.push_back(pair.second + utf8_to_utf32("</w>"));
+        }
+        for (const auto& merge : merge_pairs) {
+            vocab.push_back(merge.first + merge.second);
+        }
+        vocab.push_back(utf8_to_utf32("<|startoftext|>"));
+        vocab.push_back(utf8_to_utf32("<|endoftext|>"));
+        LOG_DEBUG("vocab size: %llu", vocab.size());
+        int i = 0;
+        for (const auto& token : vocab) {
+            encoder[token] = i++;
+        }
+
+        int rank = 0;
+        for (const auto& merge : merge_pairs) {
+            bpe_ranks[merge] = rank++;
+        }
+    };
+
+    std::u32string bpe(std::u32string token) {
+        std::u32string word = token + utf8_to_utf32("</w>");
         if (encoder.find(word) != encoder.end()) {
             return word;
         } else if (encoder.find(token) != encoder.end()) {
             return token;
         }
-        return UNK_TOKEN;
-    }
-
-    void add_token(std::string token, int32_t token_id) {
-        encoder[token] = token_id;
+        return utf8_to_utf32(UNK_TOKEN);
     }
 
     std::vector<int> tokenize(std::string text, size_t max_length = 0, bool padding = false) {
@@ -571,13 +641,25 @@ class CLIPTokenizer {
         std::vector<std::string> token_strs;
         while (std::regex_search(str, matches, pat)) {
             for (auto& token : matches) {
-                std::istringstream iss(bpe(token));
-                std::vector<std::string> tokens{std::istream_iterator<std::string>{iss},
-                                                std::istream_iterator<std::string>{}};
-                for (const auto& bpe_token : tokens) {
-                    bpe_tokens.push_back(encoder[bpe_token]);
-                    token_strs.push_back(bpe_token);
+                std::string token_str = token.str();
+                std::u32string utf32_token;
+                for (int i = 0; i < token_str.length(); i++) {
+                    char b = token_str[i];
+                    utf32_token += byte_encoder[b];
                 }
+                auto bpe_strs = bpe(utf32_token);
+                size_t start  = 0;
+                size_t pos;
+                while ((pos = bpe_strs.find(' ', start)) != std::u32string::npos) {
+                    auto bpe_str = bpe_strs.substr(start, pos - start);
+                    bpe_tokens.push_back(encoder[bpe_str]);
+                    token_strs.push_back(utf32_to_utf8(bpe_str));
+
+                    start = pos + 1;
+                }
+                auto bpe_str = bpe_strs.substr(start, bpe_strs.size() - start);
+                bpe_tokens.push_back(encoder[bpe_str]);
+                token_strs.push_back(utf32_to_utf8(bpe_str));
             }
             str = matches.suffix();
         }
@@ -4323,15 +4405,14 @@ class StableDiffusionGGML {
         LOG_INFO("Stable Diffusion weight type: %s", ggml_type_name(model_data_type));
 
         LOG_DEBUG("loading vocab");
-        auto add_token = [&](const std::string& token, int32_t token_id) {
-            cond_stage_model.tokenizer.add_token(token, token_id);
-        };
-        bool success = model_loader.load_vocab(add_token);
-        if (!success) {
-            LOG_ERROR("get vocab from file failed: '%s'", model_path.c_str());
+        std::string merges_utf8_str = model_loader.load_merges();
+        if (merges_utf8_str.size() == 0) {
+            LOG_ERROR("get merges failed: '%s'", model_path.c_str());
             return false;
         }
 
+        cond_stage_model.tokenizer.load_from_merges(merges_utf8_str);
+
         // create the ggml context for network params
         LOG_DEBUG("ggml tensor size = %d bytes", (int)sizeof(ggml_tensor));
 
@@ -4431,7 +4512,7 @@ class StableDiffusionGGML {
 
         // print_ggml_tensor(alphas_cumprod_tensor);
 
-        success = model_loader.load_tensors(on_new_tensor_cb);
+        bool success = model_loader.load_tensors(on_new_tensor_cb);
         if (!success) {
             LOG_ERROR("load tensors from file failed");
             ggml_free(ctx);

util.cpp

@@ -1,7 +1,9 @@
 #include "util.h"
 
 #include <stdarg.h>
+#include <codecvt>
 #include <fstream>
+#include <locale>
 #include <thread>
 #include <unordered_set>
 #include <vector>
@@ -119,6 +121,21 @@ int32_t get_num_physical_cores() {
     return n_threads > 0 ? (n_threads <= 4 ? n_threads : n_threads / 2) : 4;
 }
 
+std::u32string utf8_to_utf32(const std::string& utf8_str) {
+    std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> converter;
+    return converter.from_bytes(utf8_str);
+}
+
+std::string utf32_to_utf8(const std::u32string& utf32_str) {
+    std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> converter;
+    return converter.to_bytes(utf32_str);
+}
+
+std::u32string unicode_value_to_utf32(int unicode_value) {
+    std::u32string utf32_string = {static_cast<char32_t>(unicode_value)};
+    return utf32_string;
+}
+
 std::string basename(const std::string& path) {
     size_t pos = path.find_last_of('/');
     if (pos != std::string::npos) {

util.h

@@ -14,6 +14,10 @@ void replace_all_chars(std::string& str, char target, char replacement);
 bool file_exists(const std::string& filename);
 bool is_directory(const std::string& path);
 
+std::u32string utf8_to_utf32(const std::string& utf8_str);
+std::string utf32_to_utf8(const std::u32string& utf32_str);
+std::u32string unicode_value_to_utf32(int unicode_value);
+
 std::string basename(const std::string& path);
 
 std::string path_join(const std::string& p1, const std::string& p2);