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

Don't let remaining_tokens get larger than the context.

Browse source
This commit is contained in:
Johnman 2023-03-19 23:38:52 +01:00
parent dd459b604f
commit fbff268461
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
main.cpp
View file

@ -1054,7 +1054,7 @@ int main(int argc, char ** argv) {
embd_inp.insert(embd_inp.end(), inp_sfx.begin(), inp_sfx.end());
}
remaining_tokens = params.n_predict - line_inp.size();
remaining_tokens = std::min(params.n_predict, model.hparams.n_ctx - (int)embd_inp.size());
input_noecho = true; // do not echo this again
is_interacting = false;
@ -1073,7 +1073,7 @@ int main(int argc, char ** argv) {
// In interactive mode, respect the maximum number of tokens and drop back to user input when reached.
if (params.interactive && remaining_tokens <= 0) {
remaining_tokens = params.n_predict;
remaining_tokens = std::min(params.n_predict, model.hparams.n_ctx - (int)embd_inp.size()); // Will get overriden when we get to user input, but we need it to not be 0 when this iteration ends.
is_interacting = true;
}
}