Struct CalibrationDataset 

pub struct CalibrationDataset {
    pub samples: Vec<Vec<f32>>,
    pub shape: Vec<usize>,
}

A collection of FP32 calibration samples used for range estimation.

Fields

samples: Vec<Vec<f32>>

Individual samples, each flattened to match shape.

shape: Vec<usize>

Shape of a single sample (excluding batch dimension).

Implementations

impl CalibrationDataset

pub fn from_numpy(path: impl AsRef<Path>) -> Result<Self>

Load calibration samples from a NumPy .npy file.

The array must be at least 2-dimensional [batch, ...].

Requires the calibration feature (enabled by default).

Errors

Returns QuantizeError::Calibration if the file is missing, not .npy, or has an invalid shape.

pub fn random( shape: Vec<usize>, num_samples: usize, range: (f32, f32), ) -> Result<Self>

Generate random calibration samples uniformly distributed in range.

Errors

Returns QuantizeError::Calibration if shape is empty, num_samples is 0, or the range is invalid.

pub fn from_samples(samples: Vec<Vec<f32>>, shape: Vec<usize>) -> Result<Self>

Create a dataset from pre-existing sample vectors.

Errors

Returns QuantizeError::Calibration if samples is empty or any sample has the wrong length for the given shape.

pub fn sample_shape(&self) -> &[usize]

Shape of a single sample (excluding batch dimension).

pub fn len(&self) -> usize

Number of samples in the dataset.

pub fn is_empty(&self) -> bool

Whether the dataset contains no samples.

Trait Implementations

impl Clone for CalibrationDataset

fn clone(&self) -> CalibrationDataset

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CalibrationDataset

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

Formats the value using the given formatter.