nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 10 Packages Projects Releases Wiki Activity

spec: support eagle3 for qwen3.5 & 3.6 (#24593)

Browse Source 
* spec: support qwen3.5 & 3.6 eagle3 draft

* eagle3: Add deferred boundary checkpoints restore support for hybrid models

* apply suggestions

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* spec: adapt to API change

* spec: fix naming

* cont : add TODO

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Ruixiang Wang
2026-06-19 13:08:50 +03:00
committed by GitHub
co-authored by Georgi Gerganov
parent 159d093a43
commit b14e3fb90c
7 changed files with 92 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/common.h
+5 -1
View File
@@ -363,7 +363,7 @@ struct common_params_speculative {
uint32_t need_n_rs_seq() const {
bool needs_rs_seq = std::any_of(types.begin(), types.end(), [&](auto t) {
return t == COMMON_SPECULATIVE_TYPE_DRAFT_MTP;
return t == COMMON_SPECULATIVE_TYPE_DRAFT_MTP || t == COMMON_SPECULATIVE_TYPE_DRAFT_EAGLE3;
});
return needs_rs_seq ? draft.n_max : 0u;
@@ -1065,6 +1065,10 @@ struct common_prompt_checkpoint {
std::vector<uint8_t> data_tgt;
std::vector<uint8_t> data_dft;
// (optional) speculative-decoding implementation state stashed with the checkpoint
// (e.g. eagle3's deferred-boundary g_embd row)
std::vector<uint8_t> data_spec;
size_t size() const;
bool empty() const;