Function generate_column_mask 

pub fn generate_column_mask(
    scores: &Tensor,
    target_sparsity: f32,
) -> Result<SparsityMask, PruningError>

Generate a column sparsity mask.

Arguments

• scores - 2D importance scores tensor
• target_sparsity - Fraction of columns to prune

Returns

Mask where entire columns are either kept (1s) or pruned (0s).

Column sparsity is equivalent to pruning input features.