🤖 Xiangqi-RIM 1 Billion FEN NNUE Model Weights (XRNN v1)

Binary model weights (nnue_weights_1b_latest.bin) trained on 1 BILLION FENs using 100% Native Rust Engine self-play evaluations and PyTorch CUDA acceleration.

🧠 Architecture Specification (XRNN v1)

• Feature Transformer: HalfKAv2_hm ($65,536 \times 256 \to 512 \to 32 \to 1$)
• File Size: 33,571,504 Bytes (32.02 MB)
• Format: Quantized Int16 / Int8 XRNN v1 Binary Matrix

⚡ Loading in Rust Engine

use xiangrust::eval::nnue::Nnue;

let nnue = Nnue::load("data/nnue_weights_1b_latest.bin").expect("Loaded NNUE weights");