MCPcopy Create free account
hub / github.com/M4THYOU/TokenDagger / MistralTokenizer

Class MistralTokenizer

src/mistral_main.cpp:58–74  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

56}
57
58struct MistralTokenizer {
59 TokenizerConfig config;
60 std::vector<VocabItem> vocab;
61 std::unique_ptr<tiktoken::CoreBPE> bpe;
62
63 std::vector<int> encode(const std::string& prompt) const {
64 std::vector<int> tokens = {MistralSpecialTokens::BOS, MistralSpecialTokens::BEGIN_INST};
65 if (bpe) {
66 auto result = bpe->encode_ordinary(prompt);
67 for (int token : result) {
68 tokens.push_back(token + config.default_num_special_tokens);
69 }
70 }
71 tokens.push_back(MistralSpecialTokens::END_INST);
72 return tokens;
73 }
74};
75
76NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(TokenizerConfig, pattern, num_vocab_tokens, default_vocab_size, default_num_special_tokens, version);
77NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(VocabItem, rank, token_bytes, token_str);

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected