Build llama.cpp fork, verify PolarQuant, benchmark PPL/perf/memory. Target: this week.
Custom Ollama build, full test matrix, MacBook deployment, John quality review.
Per-layer profiles, QJL residual correction, upstream watch.
Build llama.cpp fork, verify PolarQuant, benchmark PPL/perf/memory. Target: this week.
Custom Ollama build, full test matrix, MacBook deployment, John quality review.
Per-layer profiles, QJL residual correction, upstream watch.