Enum ModelInputs 

pub enum ModelInputs {
    Text {
        input_ids: Tensor,
        attention_mask: Option<Tensor>,
        position_ids: Option<Tensor>,
    },
    Image {
        pixel_values: Tensor,
        image_mask: Option<Tensor>,
    },
    Multimodal {
        input_ids: Tensor,
        pixel_values: Option<Tensor>,
        attention_mask: Option<Tensor>,
        image_mask: Option<Tensor>,
    },
    Audio {
        input_features: Tensor,
        attention_mask: Option<Tensor>,
    },
}

Unified model inputs - supports all model types

Variants

Text

Text-only inputs (most language models)

Fields

input_ids: Tensor

attention_mask: Option<Tensor>

position_ids: Option<Tensor>

Image

Image-only inputs (vision models)

Fields

pixel_values: Tensor

image_mask: Option<Tensor>

Multimodal

Multimodal inputs (vision-language models)

Fields

input_ids: Tensor

pixel_values: Option<Tensor>

attention_mask: Option<Tensor>

image_mask: Option<Tensor>

Audio

Audio inputs (speech models)

Fields

input_features: Tensor

attention_mask: Option<Tensor>

Implementations

impl ModelInputs

Utility functions for model inputs/outputs

pub fn text(input_ids: Tensor) -> Self

Create text inputs

pub fn text_with_mask(input_ids: Tensor, attention_mask: Tensor) -> Self

Create text inputs with attention mask

pub fn image(pixel_values: Tensor) -> Self

Create image inputs

pub fn multimodal(input_ids: Tensor, pixel_values: Option<Tensor>) -> Self

Create multimodal inputs

pub fn batch_size(&self) -> usize

Get batch size

pub fn sequence_length(&self) -> Option<usize>

Get sequence length (for text inputs)

Trait Implementations

impl Clone for ModelInputs

fn clone(&self) -> ModelInputs

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelInputs

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

Formats the value using the given formatter.