Struct QuantizedActivations 

pub struct QuantizedActivations {
    pub data: Vec<i8>,
    pub scales: Vec<f32>,
    pub shape: Vec<usize>,
    pub per_token: bool,
}

Quantized activations with per-token scales.

Fields

data: Vec<i8>

Quantized INT8 values (stored as i8 in a Vec).

scales: Vec<f32>

Per-token scale factors. Shape depends on per_token: [batch, seq_len] or [batch].

shape: Vec<usize>

Original shape [batch, seq_len, hidden_dim].

per_token: bool

Whether per-token scaling was used.

Implementations

impl QuantizedActivations

pub fn batch_size(&self) -> usize

Get the batch size.

pub fn seq_len(&self) -> usize

Get the sequence length.

pub fn hidden_dim(&self) -> usize

Get the hidden dimension.

pub fn numel(&self) -> usize

Get the total number of elements.

Trait Implementations

impl Clone for QuantizedActivations

fn clone(&self) -> QuantizedActivations

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizedActivations

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for QuantizedActivations

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for QuantizedActivations

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.