Skip to content

llama : restore prefix space in llama tokenizer #4081

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Nov 15, 2023
Merged

llama : restore prefix space in llama tokenizer #4081

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
735ffe3
Revert "dont add space when using special tokens"
cebtenzzre Nov 14, 2023
5e89942
do not add space prefix if the first token is special
cebtenzzre Nov 15, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6283,7 +6283,10 @@ static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab &
// by modifying llm_tokenizer_x to operate with string offsets like pre-tokenizer
// and passing 'add space prefix' as bool argument
//
auto raw_text = (special ? "" : " ") + fragment.raw_text.substr(fragment.offset, fragment.length);
auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);
if (&fragment == &fragment_buffer.front()) {
raw_text = " " + raw_text; // prefix with space if the first token is not special
}

#ifdef PRETOKENIZERDEBUG
fprintf(stderr,"TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
Expand Down