# Fast-WAM Task 6

Independent GELLO specialist initialized from the pinned Fast-WAM Base LIBERO checkpoint. This is a LoRA adapter plus newly trained pose10 input/output projections, not a standalone base model.

Use this folder with the repository's `inference/serve_fastwam.py`, original base and VAE assets, this folder's normalization and exact `prompt.txt`. Input is the current external RGB image and measured base-frame flange pose10. Output is 32 absolute next-achieved flange pose10 targets, with commanded aperture, at 15 Hz target spacing. The server predicts actions without generating future video.

`INFERENCE_REPORT.json` contains the fixed forty held-out-window diagnostics and the final-weight portable/native/HTTP comparison. These checks do not establish physical task success. See the repository's `INFERENCE.md` for installation and client usage.