Tag
A user seeks clarification on the relation between MTP (Multi-Token Prediction) and QAT (Quantization-Aware Training) in llama.cpp, particularly regarding GGUF compatibility for the Gemma4 model and the new QAT string in filenames.