tpt-gpu-model-optimizer
Hardware-aware LLM model optimizer — quantization, surgical pruning, and TPTF export for TPT GPU targets.
Overview
tpt-gpu-model-optimizer loads GGUF models from the shared ~/.tpt/models/ registry, applies quantization (Q4, Q8, FP16) and optional layer pruning, and exports the result in a format ready for tpt-gpu-runtime inference. It uses AI-guided heuristics (via tpt-gpu-shared) to suggest optimal quantization strategies per target hardware.
Installation
Usage
License
Dual-licensed under MIT or Apache 2.0 WITH LLVM-exception. See LICENSE-MIT / LICENSE-APACHE for details.