Skip to content

LLM inference in C/C++

C++ ggml

server,common : fix the GCC 12 stringop-overread false positive (again) (#29325)

Adrien Gallouët2b70583997bb
diff --git a/common/speculative.cpp b/common/speculative.cpp
index 77dfe9535f..6fdfa4dc33 100644
--- a/common/speculative.cpp
+++ b/common/speculative.cpp
@@ -228,9 +228,7 @@ struct common_speculative_impl_draft_simple : public common_speculative_impl {
common_params_sampling params;
params.no_perf = false;
params.top_k = 10;
- params.samplers = {
- COMMON_SAMPLER_TYPE_TOP_K,
- };
+ params.samplers.assign(1, COMMON_SAMPLER_TYPE_TOP_K);
smpl.reset(common_sampler_init(llama_get_model(ctx_dft), params));
}
diff --git a/tools/server/server-context.cpp b/tools/server/server-context.cpp
index 274af9b05b..e95fb63ab3 100644
--- a/tools/server/server-context.cpp
+++ b/tools/server/server-context.cpp
@@ -2060,7 +2060,7 @@ private:
res->is_begin = true;
} else {
res->content = tkn.text_to_send;
- res->tokens = { tkn.tok };
+ res->tokens.assign(1, tkn.tok);
}
res->n_decoded = slot.stats.n_gen;