File tree Expand file tree Collapse file tree 3 files changed +27
-16
lines changed
include/pytorch/tokenizers Expand file tree Collapse file tree 3 files changed +27
-16
lines changed Original file line number Diff line number Diff line change 2525#include < pytorch/tokenizers/string_integer_map.h>
2626#include < pytorch/tokenizers/tokenizer.h>
2727
28+ #include " re2/re2.h"
29+
2830namespace tokenizers {
2931namespace detail {
3032
@@ -104,6 +106,25 @@ static Result<TokenMap> buildTokenMap(
104106 return buildTokenMap (std::move (pairs));
105107}
106108
109+ static Result<std::unique_ptr<IRegex>> build_special_token_regex (
110+ const TokenMap& special_token_map) {
111+ std::string special_pattern;
112+ const std::size_t count = special_token_map.size ();
113+
114+ for (std::size_t i = 0 ; i < count; ++i) {
115+ const auto & [token, _] = special_token_map.getElement (i);
116+ if (!special_pattern.empty ()) {
117+ special_pattern += " |" ;
118+ }
119+ special_pattern += re2::RE2::QuoteMeta (std::string (token));
120+ }
121+
122+ if (special_pattern.empty ()) {
123+ return static_cast <std::unique_ptr<IRegex>>(nullptr );
124+ }
125+ return create_regex (special_pattern);
126+ }
127+
107128class BPETokenizerBase : public Tokenizer {
108129 public:
109130 Result<std::vector<uint64_t >>
Original file line number Diff line number Diff line change @@ -69,6 +69,11 @@ Error HFTokenizer::load(const std::string& path) {
6969 special_tokens,
7070 [](const auto & it) -> std::string { return it.at (" content" ); },
7171 [](const auto & it) -> std::uint64_t { return it.at (" id" ); }));
72+
73+ // Create special token regex to help later with encoding.
74+ special_token_regex_ = TK_UNWRAP (detail::build_special_token_regex (special_token_map));
75+
76+ // Store for future use.
7277 special_token_map_.emplace (std::move (special_token_map));
7378 } catch (const json::out_of_range& e) {
7479 fprintf (stderr, " Could not parse special tokens: %s\n " , e.what ());
Original file line number Diff line number Diff line change 3232#include < fstream>
3333#include < limits>
3434#include < unordered_set>
35- #include " re2/re2.h"
3635
3736namespace tokenizers {
3837
@@ -47,20 +46,6 @@ static Result<std::unique_ptr<IRegex>> _create_regex(
4746 return create_regex (pattern);
4847}
4948
50- static Result<std::unique_ptr<IRegex>> _build_special_token_regex (
51- const std::vector<std::pair<std::string, std::uint64_t >>& special_encoder) {
52- std::string special_pattern;
53- for (const auto & ele : special_encoder) {
54- if (!special_pattern.empty ()) {
55- special_pattern += " |" ;
56- }
57- special_pattern += re2::RE2::QuoteMeta (ele.first );
58- }
59- if (special_pattern.empty ()) {
60- return static_cast <std::unique_ptr<IRegex>>(nullptr );
61- }
62- return _create_regex (special_pattern);
63- }
6449
6550static Result<std::pair<std::string, uint64_t >> _parse (
6651 const std::string& line) {
@@ -153,7 +138,7 @@ Error Tiktoken::load(const std::string& path) {
153138
154139 _regex = TK_UNWRAP (_create_regex (_pattern));
155140 special_token_regex_ =
156- TK_UNWRAP (_build_special_token_regex ( special_token_map));
141+ TK_UNWRAP (detail::build_special_token_regex ( TokenMap ( special_token_map) ));
157142
158143 // initialize vocab_size, bos_tok, eos_tok
159144 vocab_size_ = token_map_->size () + special_token_map_->size ();
You can’t perform that action at this time.
0 commit comments