happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Revert "sampling : stop short if backend sampler sampled a token" 

This reverts commit 87b2719eca.
This commit is contained in:
Daniel Bevenius 2025-12-04 08:26:33 +01:00
parent c0b182f4d6
commit 10bd640aae
No known key found for this signature in database
2 changed files with 0 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/llama-graph.cpp
View File

@ -2100,7 +2100,6 @@ void llm_graph_context::build_sampling() const {
if (data.sampled != nullptr) {
res->t_sampled[seq_id] = data.sampled;
ggml_build_forward_expand(gf, data.sampled);
continue;
}
if (data.probs != nullptr) {

2
tests/test-backend-sampler.cpp
View File

@ -441,8 +441,6 @@ static void test_backend_temp_sampling(const char * model_path) {
struct llama_sampler_chain_params backend_chain_params = llama_sampler_chain_default_params();
struct llama_sampler * backend_sampler_chain = llama_sampler_chain_init(backend_chain_params);
llama_sampler_chain_add(backend_sampler_chain, llama_sampler_init_temp(temp));
llama_sampler_chain_add(backend_sampler_chain, llama_sampler_init_top_k(40));
llama_sampler_chain_add(backend_sampler_chain, llama_sampler_init_dist(18));
std::vector<llama_sampler_seq_config> backend_sampler_configs = {
{ seq_id, backend_sampler_chain },