LaunchDevelopers
Jun 22, 8:31 AM
llama.cpp adds Step 3.5/3.7 flash MTP3 support
Pull request adds multi-layer MTP support for Step 3.5/3.7 flash models, following up on earlier work. Users can try it with the latest llama.cpp build.
Pull request adds multi-layer MTP support for Step 3.5/3.7 flash models, following up on earlier work. Users can try it with the latest llama.cpp build.