rook-chess-v1

A relational chess engine for the browser: it reads the 64 squares, keeps a table of how every pair of squares relates to every other, searches with MCTS, and plays one legal move.

About 3.0 million parameters, so roughly 40% the size of mini-chessformer-v1. The difference is structural rather than a matter of scale: instead of reasoning over squares alone, it carries a table of ordered square pairs through all six layers, seeded from the rules of movement (which squares attack which, and what blocks them). Same arena package contract as the other releases.

What's in this repo

This is a browser chess engine package (chess-gpt-package-v1), not a Transformers checkpoint.

A compatible runner loads:

  • browser/manifest.json โ€” file list and hashes
  • browser/entry.js โ€” search and move picker
  • the ONNX net named in that manifest

Use

Point a chess-gpt-compatible runner at burrowdweller/rook-chess-v1. It calls loadPackage โ†’ newGame โ†’ chooseMove({ history, legalMoves }) and must return one of those legal moves.

The net is small on purpose. Strength comes from searching at move time.

Source: junisbuilding/chessdb

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support