Open Source
Support Step3.5/3.7 flash mtp3 by forforever73 · Pull Request #24340 · ggml-org/llama.cpp
A pull request (#24340) has been submitted to the ggml-org/llama.cpp repository, introducing multi-layer MTP (Memory Transfer Protocol) support for models Step3.5 and Step3.7. This enhancement builds on previous work (#23274) and aims to improve memory management and efficiency in LLaMA implementations. Practitioners can leverage this update to optimize the performance of their applications using LLaMA models with enhanced memory capabilities.
llamapull_requestgguf