nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 17 Packages Projects Releases Wiki Activity

qwen35: use post-norm hidden state for MTP (#24025)

Browse Source 
* qwen35: use post-norm hidden state for MTP

* rename pre_norm to nextn

* fix step35
This commit is contained in:
Aman Gupta
2026-06-04 01:29:09 +08:00
committed by GitHub
parent c8d6a00636
commit 166fe29492
12 changed files with 132 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/speculative.h
+2 -2
View File
@@ -59,8 +59,8 @@ bool common_speculative_process(common_speculative * spec, const llama_batch & b
// true if any implementation requires target post-norm embeddings to be extracted
bool common_speculative_need_embd(common_speculative * spec);
// true if any implementation requires target pre-norm embeddings to be extracted
bool common_speculative_need_embd_pre_norm(common_speculative * spec);
// true if any implementation requires target nextn embeddings to be extracted
bool common_speculative_need_embd_nextn(common_speculative * spec);
// generate drafts for the sequences specified with `common_speculative_get_draft_params`
void common_speculative_draft(common_speculative * spec);