{% extends "base.html" %} {% block title %}GameGame learning and native artifacts{% endblock %} {% block page_id %}gamegame_learning{% endblock %} {% block head %}{% endblock %} {% block content %}

GameGame learning and native artifacts

Shared Builder datasets and Gym, Bot learned exports, and the actual Hardware Lab practice ladder.

Native Bot, Builder and Detailer controlsNative datasets and fine-tuning

Builder datasets and Gym

The dataset generator rejudges saved Builder builds and recipes with the installed native game compiler. Detailer recipes are excluded. Gym runs the native Planner and model without placing objects. If this Builder is running, its practice waits while Gym runs, then its prior practice mode is restored and it imports the generated clean recipes. Restoring practice can resume its existing game activity. Actual pause, restoration and ingestion outcomes appear in the result.

Bot learned record

Capture the selected running Bot's actual /learned response as the desktop's learned JSON dataset and training-job baseline.

Engineering practice

Eleven training tasks and four held-out tasks run against the model and the simulated Hardware Lab. This does not connect physical hardware. Held-out answers and lessons are excluded from training exports.



Detailer artifacts
The native Detailer offers saved recipe/build artifacts and its finishing controls. Its native panel has no Gym or dataset generator.


Reviewed native source state


Graceful Stop preserves completed native results. Accepted writes use their actual result. Sign-out, expiry, or server shutdown closes the owned worker tree and discards output that has not yet been published. Changed native profiles, source code, or saved input files prevent publication.


Actual operation result
Native artifacts

{% endblock %}
{% block scripts %}{% endblock %}