Add separately calibrated RTX5090 NVFP4 transformer as one model file

Publish consolidated transformer weights; simplify model repository

Complete calibration Docker context and clarify reproducible cache setup

Finalize tested RunPod wrapper, clean Docker build and reproduction instructions

Publish calibrated rank128 INT4 research model and reproducible runtime (part 2)

Publish calibrated rank128 INT4 research model and reproducible runtime

initial commit