ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-02-21 13:44:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

add llama_sampling_sample_and_accept_n to sampling

Browse Source
This commit is contained in:
T. M.
2025-07-25 03:39:59 +00:00
parent 5c96a7f619
commit 99c1ef3c01
2 changed files with 50 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
common/sampling.cpp
View File

@@ -506,3 +506,34 @@ void llama_sampling_accept(
llama_sampler_dry_accept(ctx_sampling->smpl, id);
}
}
std::vector<llama_token> llama_sampling_sample_and_accept_n(struct llama_sampling_context * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const std::vector<llama_token> & draft) {
GGML_ASSERT(idxs.size() == draft.size() + 1 && "idxs.size() must be draft.size() + 1");
std::vector<llama_token> result;
result.reserve(idxs.size());
size_t i = 0;
for (; i < draft.size(); i++) {
const llama_token id = llama_sampling_sample(gsmpl, ctx, nullptr, idxs[i]);
llama_sampling_accept(gsmpl, ctx, id, true);
result.push_back(id);
if (draft[i] != id) {
break;
}
}
if (i == draft.size()) {
const llama_token id = llama_sampling_sample(gsmpl, ctx, nullptr, idxs[i]);
llama_sampling_accept(gsmpl, ctx, id, true);
result.push_back(id);
}
return result;
}