sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-05-13 12:34:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

experiments

Browse Source
This commit is contained in:
Georgi Gerganov
2026-01-28 09:45:07 +02:00
parent 003c90352d
commit 6c8a04576e
6 changed files with 189 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tools/server/server-context.cpp
View File

@@ -2039,9 +2039,15 @@ private:
/*.params_spec.n_draft =*/ n_draft_max,
/*.params_spec.p_min =*/ slot.task->params.speculative.p_min,
};
const llama_tokens & cached_text_tokens = slot.prompt.tokens.get_text_tokens();
llama_tokens draft = common_speculative_gen_draft(slot.spec, params_spec, cached_text_tokens, slot.sampled);
if (draft.size() > 0) {
std::string tmp = common_detokenize(slot.ctx, draft);
//LOG_WRN("XXXXXX: draft: '%s'\n", tmp.c_str());
}
// add the sampled token to the batch
slot.i_batch_dft.push_back(batch.n_tokens);
common_batch_add(batch, slot.sampled, slot.prompt.tokens.pos_next(), { slot.id }, true);