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

Only use FIM middle token if it exists (#7648)

Browse source 
* Only use FIM middle if it exists

* Only use FIM middle if it exists
This commit is contained in:
Sigbjørn Skjæret 2024-06-18 14:19:45 +02:00 • committed by GitHub
parent 84f6de17f6
commit 91c188d6c2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 17 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

7
examples/server/server.cpp
View file

@ -2038,7 +2038,12 @@ struct server_context {
prefix_tokens.insert(prefix_tokens.begin(), llama_token_bos(model)); // always add BOS
prefix_tokens.insert(prefix_tokens.end(), llama_token_suffix(model));
prefix_tokens.insert(prefix_tokens.end(), suffix_tokens.begin(), suffix_tokens.end());
prefix_tokens.push_back(llama_token_middle(model));
const llama_token middle_token = llama_token_middle(model);
if (middle_token >= 0) {
prefix_tokens.push_back(middle_token);
}
prompt_tokens = prefix_tokens;
} else {
prompt_tokens = tokenize(slot.prompt, system_prompt.empty()); // add BOS if there isn't system prompt