github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-06 11:00:48 +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/gptj.cpp
View File

@@ -993,9 +993,10 @@ void GPTJ::prompt(const std::string &prompt,
gpt_vocab::id id = 0;
{
const int64_t t_start_sample_us = ggml_time_us();
const size_t n_prev_toks = std::min((size_t) promptCtx.repeat_last_n, promptCtx.tokens.size());
id = gpt_sample_top_k_top_p(d_ptr->vocab, n_vocab,
promptCtx.tokens.data() + promptCtx.n_ctx - promptCtx.n_ctx,
promptCtx.n_ctx,
promptCtx.tokens.data() + promptCtx.tokens.size() - n_prev_toks,
n_prev_toks,
promptCtx.logits,
promptCtx.top_k, promptCtx.top_p, promptCtx.temp,
promptCtx.repeat_penalty,