github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-10 12:59:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

backend: fix buffer overrun in repeat penalty code

Browse Source 
Caught with AddressSanitizer running a basic prompt test against llmodel
standalone. This fix allows ASan builds to complete a simple prompt
without illegal accesses but there are still notably several leaks.
This commit is contained in:
aaron miller
2023-05-16 23:20:08 -07:00
committed by AT
parent 26cb31c4e6
commit e6fd0a240d
3 changed files with 9 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
gpt4all-backend/llamamodel.cpp
View File

@@ -180,9 +180,10 @@ void LLamaModel::prompt(const std::string &prompt,
int32_t totalPredictions = 0;
for (int i = 0; i < promptCtx.n_predict; i++) {
// sample next token
const size_t n_prev_toks = std::min((size_t) promptCtx.repeat_last_n, promptCtx.tokens.size());
llama_token id = llama_sample_top_p_top_k(d_ptr->ctx,
promptCtx.tokens.data() + promptCtx.n_ctx - promptCtx.repeat_last_n,
promptCtx.repeat_last_n, promptCtx.top_k, promptCtx.top_p, promptCtx.temp,
promptCtx.tokens.data() + promptCtx.tokens.size() - n_prev_toks,
n_prev_toks, promptCtx.top_k, promptCtx.top_p, promptCtx.temp,
promptCtx.repeat_penalty);
// Check if the context has run out...