nicolas
/
ik_llama_opt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix rope type for 39 architectures (#2290) 

The DFLASH case was inserted into the NEOX fall-through group in
llama_rope_type(), so the 39 cases above it now return LLAMA_ROPE_TYPE_NORM
instead of LLAMA_ROPE_TYPE_NEOX.

Move DFLASH into the NORM group so it keeps its intended rope type and the
others fall through to NEOX again.
This commit is contained in:
mb8565 2026-08-11 01:09:24 -05:00 committed by GitHub
parent 5763a901de
commit c8772a8429
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/llama.cpp
View File

@ -8782,6 +8782,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
case LLM_ARCH_MISTRAL3:
case LLM_ARCH_GLM_DSA:
case LLM_ARCH_MISTRAL4:
case LLM_ARCH_DFLASH:
return LLAMA_ROPE_TYPE_NORM;
// the pairs of head values are offset by n_rot/2
@ -8824,8 +8825,6 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
case LLM_ARCH_LAGUNA:
case LLM_ARCH_GEMMA4:
case LLM_ARCH_GEMMA4_MTP:
case LLM_ARCH_DFLASH:
return LLAMA_ROPE_TYPE_NORM;
case LLM_ARCH_DFLASH_DRAFT:
case LLM_ARCH_GEMMA4_ASSISTANT:
return LLAMA_ROPE_TYPE_NEOX;