Superminaren/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-04 02:37:27 +02:00
Code Issues Packages Projects Releases Wiki Activity

convert_hf_to_gguf: support split MTP export for HY V3 (#25641)

Browse Source 
- Add a supports_mtp_export capability to ModelBase so architectures can opt
  into --mtp and --no-mtp without extending a central class allowlist.
- Enable the capability for the existing Qwen3.5/3.6 and Step3.5/3.7
  implementations, and for HY V3, whose converter already supports
  filtering the appended MTP layers.
This commit is contained in:
Thiago Padilha
2026-07-14 11:43:15 +02:00
committed by GitHub
parent ec0dbef816
commit cb489bc0fb
5 changed files with 8 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
convert_hf_to_gguf.py
+2 -4
View File
@@ -259,10 +259,8 @@ def main() -> None:
sys.exit(1)
if args.mtp or args.no_mtp:
from conversion.qwen import _Qwen35MtpMixin
from conversion.step3 import Step35Model
if not (issubclass(model_class, _Qwen35MtpMixin) or issubclass(model_class, Step35Model)):
logger.error("--mtp / --no-mtp are only supported for Qwen3.5/3.6 and Step3.5 text variants today")
if not model_class.supports_mtp_export:
logger.error("--mtp / --no-mtp are not supported for %s", model_architecture)
sys.exit(1)
if args.no_mtp:
model_class.no_mtp = True