Function quantize_weights 

pub fn quantize_weights(
    weight: &Tensor,
    config: &BitNetConfig,
) -> Result<TernaryWeight>

Quantize a weight tensor to ternary using AbsMean quantization.

Algorithm

For each group of weights:

1. Compute scale = mean(|W|)
2. Compute W_q = round(W / scale) clamped to {-1, 0, +1}

Arguments

• weight - Input weight tensor [out_features, in_features]
• config - BitNet configuration

Errors

Returns error if weight has wrong shape or quantization fails.