Trait drug::nodes::Operation

pub trait Operation: Debug {
    fn eval(&self, inputs: &[ArrayViewD<'_, f32>]) -> ArrayD<f32>;
    fn grad(
        &self,
        inputs: &[ArrayViewD<'_, f32>],
        loss: ArrayViewD<'_, f32>
    ) -> Vec<ArrayD<f32>>;
}

Represents a differentiable function in a computation graph. Operations hold their own hyperparameters but not their parameters, values or losses. Unfortunately boxed traits cannot be saved with serde. When reloaded they will be replaced by Box<arithmetic::Add> nodes. When reloading a model with custom Operations, you need to replace them manually.

Required Methods

fn eval(&self, inputs: &[ArrayViewD<'_, f32>]) -> ArrayD<f32>

Mutates Outputs based on inputs. TODO consider modifying output ArrayD in place

fn grad(
    &self,
    inputs: &[ArrayViewD<'_, f32>],
    loss: ArrayViewD<'_, f32>
) -> Vec<ArrayD<f32>>

Returns gradients of inputs wrt outputs. Note the inputs and output vectors should be the same length. TODO consider modifying output ArrayDs in place

Trait Implementations

impl Default for Box<dyn Operation>

fn default() -> Self

Returns the “default value” for a type.

impl<'erased> Serialize for dyn Operation + 'erased

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>where
    S: Serializer,

Serialize this value into the given Serde serializer.

impl<'erased> Serialize for dyn Operation + Send + 'erased

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>where
    S: Serializer,

Serialize this value into the given Serde serializer.

impl<'erased> Serialize for dyn Operation + Send + Sync + 'erased

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>where
    S: Serializer,

Serialize this value into the given Serde serializer.

impl<'erased> Serialize for dyn Operation + Sync + 'erased

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>where
    S: Serializer,

Serialize this value into the given Serde serializer.

Implementors

impl Operation for Activation

impl Operation for GlobalPool

impl Operation for Add

impl Operation for Conv

impl Operation for Embedding

impl Operation for MatMul

impl Operation for Mult