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

Pause sampling if waiting for user input.

Browse source
This commit is contained in:
Johnman 2023-03-19 16:57:02 +01:00
parent 10f1c9ed30
commit b78caa6bff
1 changed files with 24 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

46
main.cpp
View file

@ -935,35 +935,37 @@ int main(int argc, char ** argv) {
embd.clear(); embd.clear();
if (embd_inp.size() <= input_consumed) { if (embd_inp.size() <= input_consumed) {
// out of user input, sample next token if (!is_interacting) {
const float top_k = params.top_k; // out of user input, sample next token
const float top_p = params.top_p; const float top_k = params.top_k;
const float temp = params.temp; const float top_p = params.top_p;
const float repeat_penalty = params.repeat_penalty; const float temp = params.temp;
const float repeat_penalty = params.repeat_penalty;
const int n_vocab = model.hparams.n_vocab; const int n_vocab = model.hparams.n_vocab;
gpt_vocab::id id = 0; gpt_vocab::id id = 0;
{ {
const int64_t t_start_sample_us = ggml_time_us(); const int64_t t_start_sample_us = ggml_time_us();
id = llama_sample_top_p_top_k(vocab, logits.data() + (logits.size() - n_vocab), last_n_tokens, repeat_penalty, top_k, top_p, temp, rng); id = llama_sample_top_p_top_k(vocab, logits.data() + (logits.size() - n_vocab), last_n_tokens, repeat_penalty, top_k, top_p, temp, rng);
last_n_tokens.erase(last_n_tokens.begin()); last_n_tokens.erase(last_n_tokens.begin());
last_n_tokens.push_back(id); last_n_tokens.push_back(id);
t_sample_us += ggml_time_us() - t_start_sample_us; t_sample_us += ggml_time_us() - t_start_sample_us;
}
// add it to the context
embd.push_back(id);
// echo this to console
input_noecho = false;
// decrement remaining sampling budget
--remaining_tokens;
} }
// add it to the context
embd.push_back(id);
// echo this to console
input_noecho = false;
// decrement remaining sampling budget
--remaining_tokens;
} else { } else {
// some user input remains from prompt or interaction, forward it to processing // some user input remains from prompt or interaction, forward it to processing
while (embd_inp.size() > input_consumed) { while (embd_inp.size() > input_consumed) {