Skip to content
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

Jump to bottom

Fix incorrect prompt tokenization in speculative example #4025

Merged
merged 3 commits into from
Nov 20, 2023
Merged

Fix incorrect prompt tokenization in speculative example #4025

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c2d0041
Support special tokens and not adding BOS to prompt in speculative
AutonomicPerfectionist Nov 10, 2023
e778ce4
Adapt to new should_add_bos function
AutonomicPerfectionist Nov 18, 2023
9cfc5e2
Ensure tgt and dft have same add_bos setting
AutonomicPerfectionist Nov 18, 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
17 changes: 15 additions & 2 deletions examples/speculative/speculative.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,9 +94,22 @@ int main(int argc, char ** argv) {
}
}

// tokenize the prompt

// Tokenize the prompt
const bool add_bos_tgt = llama_should_add_bos_token(model_tgt);
LOG("add_bos tgt: %d\n", add_bos_tgt);

const bool add_bos_dft = llama_should_add_bos_token(model_dft);
LOG("add_bos dft: %d\n", add_bos_dft);

if (add_bos_tgt != add_bos_dft) {
fprintf(stderr, "%s: error: draft model add_bos must match target model to use speculation but ", __func__);
fprintf(stderr, "add_bos_dft = %d while add_bos_tgt = %d\n", add_bos_dft, add_bos_tgt);
return 1;
}

std::vector<llama_token> inp;
inp = ::llama_tokenize(ctx_tgt, params.prompt, true);
inp = ::llama_tokenize(ctx_tgt, params.prompt, add_bos_tgt, true);

const int max_context_size = llama_n_ctx(ctx_tgt);
const int max_tokens_list_size = max_context_size - 4;
Expand Down
Loading