vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix deepseek deseret regex

Browse source 
On windows compiled with gcc the c++ regex library failed to handle
the characters
This commit is contained in:
Daniel Hiltgen 2024-10-25 16:25:18 -07:00
parent cc2983d375
commit 439ca3bd04
2 changed files with 22 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/llama-vocab.cpp
View file

@ -389,7 +389,7 @@ struct llm_tokenizer_bpe : llm_tokenizer {
case LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM: case LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM:
regex_exprs = { regex_exprs = {
"[\r\n]", "[\r\n]",
"\\s?[A-Za-zµÀ-ÖØ-öø-ƺƼ-ƿDŽ-ʓʕ-ʯͰ-ͳͶͷͻ-ͽͿΆΈ-ΊΌΎ-ΡΣ-ϵϷ-ҁҊ-ԯԱ-ՖႠ-ჅᎠ-Ᏽᏸ-ᏽᲐ-ᲺᲽ-Ჿᴀ-ᴫᵫ-ᵷᵹ-ᶚḀ-ἕἘ-Ἕἠ-ὅὈ-Ὅὐ-ὗὙὛὝὟ-ώᾀ-ᾴᾶ-ᾼιῂ-ῄῆ-ῌῐ-ΐῖ-Ίῠ-Ῥῲ-ῴῶ-ῼℂℇℊ-ℓℕℙ-ℝℤΩℨK-ℭℯ-ℴℹℼ-ℿⅅ-ⅉⅎↃↄⰀ-ⱻⱾ-ⳤⳫ-ⳮⳲⳳꙀ-ꙭꚀ-ꚛꜢ-ꝯꝱ-ꞇꞋ-ꞎꭰ-ꮿff-stﬓ-ﬗＡ-Ｚａ-ｚ𐐀-𐑏𐒰-𐓓𐓘-𐓻𐲀-𐲲𐳀-𐳲𑢠-𑣟𞤀-𞥃]+", "\\s?[A-Za-zµÀ-ÖØ-öø-ƺƼ-ƿDŽ-ʓʕ-ʯͰ-ͳͶͷͻ-ͽͿΆΈ-ΊΌΎ-ΡΣ-ϵϷ-ҁҊ-ԯԱ-ՖႠ-ჅᎠ-Ᏽᏸ-ᏽᲐ-ᲺᲽ-Ჿᴀ-ᴫᵫ-ᵷᵹ-ᶚḀ-ἕἘ-Ἕἠ-ὅὈ-Ὅὐ-ὗὙὛὝὟ-ώᾀ-ᾴᾶ-ᾼιῂ-ῄῆ-ῌῐ-ΐῖ-Ίῠ-Ῥῲ-ῴῶ-ῼℂℇℊ-ℓℕℙ-ℝℤΩℨK-ℭℯ-ℴℹℼ-ℿⅅ-ⅉⅎↃↄⰀ-ⱻⱾ-ⳤⳫ-ⳮⳲⳳꙀ-ꙭꚀ-ꚛꜢ-ꝯꝱ-ꞇꞋ-ꞎꭰ-ꮿff-stﬓ-ﬗＡ-Ｚａ-ｚ\U00010400-\U0001044f𐒰-𐓓𐓘-𐓻𐲀-𐲲𐳀-𐳲𑢠-𑣟𞤀-𞥃]+",
"\\s?[!-/:-~！-／：-～‘-‟　-。]+", "\\s?[!-/:-~！-／：-～‘-‟　-。]+",
"\\s+$", "\\s+$",
"[一-龥ࠀ-一가-퟿]+", "[一-龥ࠀ-一가-퟿]+",

21
src/unicode.cpp
View file

@ -2,6 +2,11 @@
#define _SILENCE_CXX17_CODECVT_HEADER_DEPRECATION_WARNING #define _SILENCE_CXX17_CODECVT_HEADER_DEPRECATION_WARNING
#endif #endif
#if defined(_WIN32)
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#endif
#include "unicode.h" #include "unicode.h"
#include "unicode-data.h" #include "unicode-data.h"
@ -201,8 +206,24 @@ static std::unordered_map<std::string, uint8_t> unicode_utf8_to_byte_map() {
} }
static inline std::wstring unicode_wstring_from_utf8(const std::string & s) { static inline std::wstring unicode_wstring_from_utf8(const std::string & s) {
#ifdef _WIN32
int wlen = MultiByteToWideChar(CP_UTF8, 0, s.c_str(), -1, NULL, 0);
if (!wlen) {
throw std::invalid_argument("failed to convert regex");
}
wchar_t * wbuf = (wchar_t *) malloc(wlen * sizeof(wchar_t));
wlen = MultiByteToWideChar(CP_UTF8, 0, s.c_str(), -1, wbuf, wlen);
if (!wlen) {
free(wbuf);
throw std::invalid_argument("failed to convert regex");
}
std::wstring ret = std::wstring(wbuf);
free(wbuf);
return ret;
#else
std::wstring_convert<std::codecvt_utf8<wchar_t>> conv; std::wstring_convert<std::codecvt_utf8<wchar_t>> conv;
return conv.from_bytes(s); return conv.from_bytes(s);
#endif
} }
static std::vector<std::string> unicode_byte_encoding_process(const std::vector<std::string> & bpe_words) { static std::vector<std::string> unicode_byte_encoding_process(const std::vector<std::string> & bpe_words) {