Struct TabDptConfig 

pub struct TabDptConfig {

    pub dim: usize,
    pub n_layers: usize,
    pub n_heads: usize,
    pub ff_dim: usize,
    pub y_encoder_dim: usize,
    pub max_num_classes: usize,
    pub regression_bin_count: usize,
    pub regression_bin_min: f32,
    pub regression_bin_max: f32,
    pub max_num_features: usize,
    pub base_len: usize,
    pub max_len: usize,
    pub n_thinking_rows: usize,
}

TabDPT model configuration. One checkpoint (Layer6/TabDPT) serves both classification and regression — the head produces max_num_classes + regression_bin_count outputs; callers slice whichever half they need.

Fields

dim: usize

n_layers: usize

n_heads: usize

ff_dim: usize

y_encoder_dim: usize

max_num_classes: usize

regression_bin_count: usize

regression_bin_min: f32

regression_bin_max: f32

max_num_features: usize

base_len: usize

max_len: usize

n_thinking_rows: usize

Implementations

impl TabDptConfig

pub fn default_v1_2() -> Self

Matches Layer6/TabDPT’s tabdpt1_2.safetensors embedded config exactly.

pub fn kappa(&self) -> Option<f64>

kappa = (sqrt(head_dim) - 1) / ln(max_len / base_len), used by every layer’s attention temperature scaling. None (scaling disabled) when base_len == max_len.

Trait Implementations

impl Clone for TabDptConfig

fn clone(&self) -> TabDptConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for TabDptConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.