ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Port universal assisted decoding to llama-server (#699)

Browse Source 
* port universal assisted decoding to server

* fix calls

* fix LOG_INFO

* fix llama_detokenize call

* use emplace_back
This commit is contained in:
g2mt
2025-08-17 23:22:23 -07:00
committed by GitHub
parent 6b2c84b099
commit 06bed7e01b
5 changed files with 160 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
common/speculative.h
View File

@@ -13,10 +13,17 @@ struct llama_speculative_params {
float p_min = 0.75f; // min probability required to accept a token in the draft
};
struct llama_speculative * llama_speculative_init(struct llama_context * ctx_dft);
struct llama_speculative * llama_speculative_init(
struct llama_context * ctx_tgt,
struct llama_context * ctx_dft
);
void llama_speculative_free(struct llama_speculative * spec);
void llama_speculative_add_replacement_tgt_dft(
struct llama_speculative * spec,
const char *source, const char *dest);
bool llama_speculative_are_compatible(
const struct llama_context * ctx_tgt,
const struct llama_context * ctx_dft);