github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-10-08 13:39:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Don't repeat the prompt in the response.

Browse Source
This commit is contained in:
Adam Treat
2023-04-09 01:11:52 -04:00
parent e35fcd038c
commit c430ed12c6
2 changed files with 9 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
gptj.cpp
View File

@@ -700,6 +700,7 @@ void GPTJ::prompt(const std::string &prompt, std::function<bool(const std::strin
n_predict = std::min(n_predict, d_ptr->model.hparams.n_ctx - (int) embd_inp.size());
std::vector<gpt_vocab::id> embd;
std::vector<gpt_vocab::id> resp;
// determine the required inference memory per token:
size_t mem_per_token = 0;
@@ -720,6 +721,7 @@ void GPTJ::prompt(const std::string &prompt, std::function<bool(const std::strin
n_past += embd.size();
embd.clear();
resp.clear();
if (i >= embd_inp.size()) {
// sample next token
@@ -738,6 +740,7 @@ void GPTJ::prompt(const std::string &prompt, std::function<bool(const std::strin
// add it to the context
embd.push_back(id);
resp.push_back(id);
} else {
// if here, it means we are still processing the input prompt
for (int k = i; k < embd_inp.size(); k++) {
@@ -750,7 +753,7 @@ void GPTJ::prompt(const std::string &prompt, std::function<bool(const std::strin
}
// display text
for (auto id : embd) {
for (auto id : resp) {
if (!response(d_ptr->vocab.id_to_token[id]))
goto stop_generating;
}
@@ -762,7 +765,7 @@ void GPTJ::prompt(const std::string &prompt, std::function<bool(const std::strin
}
stop_generating:
#if 1
#if 0
// report timing
{
const int64_t t_main_end_us = ggml_time_us();