🎨 Clean up axis operators, move them to crate root

src/axis.rs

@@ -1,46 +1,64 @@
 use std::borrow::Borrow;
+use std::fmt::Debug;
 
 use nalgebra::{Const, DVector, Dyn, Scalar, VecStorage};
 
-use crate::{err::NeuraAxisErr, prelude::NeuraShape};
+use crate::err::NeuraAxisErr;
+use crate::prelude::NeuraShape;
 
-// TODO: create a NeuraAxis trait
+pub trait NeuraAxisBase: Clone + Debug + 'static {
+    type Err: Debug;
 
-#[derive(Clone, Copy, Debug)]
+    fn shape(&self, input_shapes: &[NeuraShape]) -> Result<NeuraShape, Self::Err>;
-pub struct NeuraAxisAppend;
+}
 
-pub trait NeuraCombineInputs<T> {
+/// Axis operators take in a set of inputs, and combine them together into one output,
-    type Combined;
+/// which is then usually fed to a layer.
+pub trait NeuraAxis<Input>: NeuraAxisBase {
+    type Combined: 'static;
 
-    fn combine(&self, inputs: Vec<impl Borrow<T>>) -> Self::Combined;
+    fn combine(&self, inputs: &[impl Borrow<Input>]) -> Self::Combined;
-}
 
-pub trait NeuraSplitInputs<T>: NeuraCombineInputs<T> {
+    fn split(&self, combined: &Self::Combined, input_shapes: &[NeuraShape]) -> Vec<Input>;
-    fn split(&self, combined: &Self::Combined, input_shapes: &[NeuraShape]) -> Vec<T>;
 }
 
-impl<F: Clone> NeuraCombineInputs<DVector<F>> for NeuraAxisAppend {
+/// An axis operator that
-    type Combined = DVector<F>;
+#[derive(Clone, Debug)]
+pub struct NeuraAxisDefault;
 
-    fn combine(&self, inputs: Vec<impl Borrow<DVector<F>>>) -> Self::Combined {
+impl NeuraAxisBase for NeuraAxisDefault {
-        assert!(inputs.len() > 0);
+    type Err = NeuraAxisErr;
-        let mut res = Vec::with_capacity(inputs.iter().map(|vec| vec.borrow().len()).sum());
 
-        for input in inputs {
+    fn shape(&self, inputs: &[NeuraShape]) -> Result<NeuraShape, NeuraAxisErr> {
-            for x in input.borrow().iter() {
+        if inputs.len() != 1 {
-                res.push(x.clone());
+            Err(NeuraAxisErr::InvalidAmount(inputs.len(), 1, Some(1)))
+        } else {
+            Ok(*inputs[0].borrow())
+        }
     }
 }
 
-        DVector::from_data(VecStorage::new(Dyn(res.len()), Const as Const<1>, res))
+impl<Data: Clone + 'static> NeuraAxis<Data> for NeuraAxisDefault {
+    type Combined = Data;
+
+    fn combine(&self, inputs: &[impl Borrow<Data>]) -> Self::Combined {
+        assert!(inputs.len() == 1);
+
+        inputs[0].borrow().clone()
+    }
+
+    fn split(&self, combined: &Self::Combined, _input_shapes: &[NeuraShape]) -> Vec<Data> {
+        vec![combined.clone()]
     }
 }
 
-// TODO: use another trait for combining NeuraShape, or make it another member of the trait
+#[derive(Clone, Copy, Debug)]
-impl NeuraCombineInputs<NeuraShape> for NeuraAxisAppend {
+pub struct NeuraAxisAppend;
-    type Combined = Result<NeuraShape, NeuraAxisErr>;
 
-    fn combine(&self, inputs: Vec<impl Borrow<NeuraShape>>) -> Self::Combined {
+impl NeuraAxisBase for NeuraAxisAppend {
+    type Err = NeuraAxisErr;
+
+    fn shape(&self, inputs: &[NeuraShape]) -> Result<NeuraShape, NeuraAxisErr> {
         let mut inputs = inputs.into_iter().map(|x| *x.borrow());
         if let Some(mut res) = inputs.next() {
             for operand in inputs {
@@ -60,7 +78,22 @@ impl NeuraCombineInputs<NeuraShape> for NeuraAxisAppend {
     }
 }
 
-impl<F: Clone + Scalar + Default> NeuraSplitInputs<DVector<F>> for NeuraAxisAppend {
+impl<F: Clone + Default + Scalar> NeuraAxis<DVector<F>> for NeuraAxisAppend {
+    type Combined = DVector<F>;
+
+    fn combine(&self, inputs: &[impl Borrow<DVector<F>>]) -> Self::Combined {
+        assert!(inputs.len() > 0);
+        let mut res = Vec::with_capacity(inputs.iter().map(|vec| vec.borrow().len()).sum());
+
+        for input in inputs {
+            for x in input.borrow().iter() {
+                res.push(x.clone());
+            }
+        }
+
+        DVector::from_data(VecStorage::new(Dyn(res.len()), Const as Const<1>, res))
+    }
+
     fn split(&self, combined: &Self::Combined, input_shapes: &[NeuraShape]) -> Vec<DVector<F>> {
         let mut result = Vec::with_capacity(input_shapes.len());
         let mut offset = 0;
@@ -83,37 +116,3 @@ impl<F: Clone + Scalar + Default> NeuraSplitInputs<DVector<F>> for NeuraAxisAppe
         result
     }
 }
-
-#[derive(Clone, Debug)]
-pub struct NeuraAxisDefault;
-
-impl<F: Clone> NeuraCombineInputs<DVector<F>> for NeuraAxisDefault {
-    type Combined = DVector<F>;
-
-    fn combine(&self, inputs: Vec<impl Borrow<DVector<F>>>) -> Self::Combined {
-        assert!(inputs.len() == 1);
-
-        inputs[0].borrow().clone()
-    }
-}
-
-impl NeuraCombineInputs<NeuraShape> for NeuraAxisDefault {
-    type Combined = Result<NeuraShape, NeuraAxisErr>;
-
-    fn combine(&self, inputs: Vec<impl Borrow<NeuraShape>>) -> Self::Combined {
-        if inputs.len() != 1 {
-            Err(NeuraAxisErr::InvalidAmount(inputs.len(), 1, Some(1)))
-        } else {
-            Ok(*inputs[0].borrow())
-        }
-    }
-}
-
-impl<Data: Clone> NeuraSplitInputs<Data> for NeuraAxisDefault
-where
-    NeuraAxisDefault: NeuraCombineInputs<Data, Combined = Data>,
-{
-    fn split(&self, combined: &Self::Combined, _input_shapes: &[NeuraShape]) -> Vec<Data> {
-        vec![combined.clone()]
-    }
-}

src/err.rs

@@ -75,10 +75,10 @@ pub enum NeuraAxisErr {
 }
 
 #[derive(Clone, Debug)]
-pub enum NeuraResidualConstructErr<LayerErr> {
+pub enum NeuraResidualConstructErr<LayerErr, AxisErr> {
     Layer(LayerErr),
     WrongConnection(isize),
-    AxisErr(NeuraAxisErr),
+    AxisErr(AxisErr),
     NoOutput,
 }
 

src/lib.rs

@@ -1,4 +1,5 @@
 pub mod algebra;
+pub mod axis;
 pub mod derivable;
 pub mod err;
 pub mod gradient_solver;

src/network/graph/from.rs

@@ -1,7 +1,4 @@
-use crate::network::{
+use crate::network::sequential::NeuraSequentialLast;
-    residual::{NeuraAxisDefault, NeuraSplitInputs},
-    sequential::NeuraSequentialLast,
-};
 
 use super::*;
 
@@ -32,7 +29,6 @@ impl<Data: Clone + 'static, Layer: NeuraLayer<Data, Output = Data>, ChildNetwork
     FromSequential<NeuraSequential<Layer, ChildNetwork>, Data> for NeuraGraph<Data>
 where
     NeuraGraph<Data>: FromSequential<ChildNetwork, Data>,
-    NeuraAxisDefault: NeuraSplitInputs<Data, Combined = Data>,
     Layer::IntermediaryRepr: 'static,
 {
     fn from_sequential_rec(

src/network/graph/mod.rs

@@ -13,11 +13,6 @@ pub use partial::NeuraGraphPartial;
 mod from;
 pub use from::FromSequential;
 
-#[deprecated]
-pub trait NeuraTrainableLayerFull<Input>: NeuraLayer<Input> {}
-
-impl<Input, T> NeuraTrainableLayerFull<Input> for T where T: NeuraLayer<Input> {}
-
 #[derive(Debug)]
 pub struct NeuraGraphNodeConstructed<Data> {
     node: Box<dyn NeuraGraphNodeEval<Data>>,
@@ -202,7 +197,7 @@ impl<Data: Clone + std::fmt::Debug + 'static> NeuraLayer<Data> for NeuraGraph<Da
 
 #[cfg(test)]
 mod test {
-    use crate::{err::NeuraGraphErr, network::residual::NeuraAxisAppend, utils::uniform_vector};
+    use crate::{axis::NeuraAxisAppend, err::NeuraGraphErr, utils::uniform_vector};
 
     use super::*;
 

src/network/graph/node.rs

@@ -3,14 +3,12 @@ use std::{any::Any, fmt::Debug};
 
 use crate::{
     algebra::NeuraDynVectorSpace,
-    err::NeuraAxisErr,
+    axis::{NeuraAxis, NeuraAxisDefault},
-    network::residual::{NeuraAxisDefault, NeuraCombineInputs, NeuraSplitInputs},
     prelude::{NeuraPartialLayer, NeuraShape},
 };
 
 use super::*;
 
-// TODO: split into two  traits
 pub trait NeuraGraphNodePartial<Data>: DynClone + Debug {
     fn inputs<'a>(&'a self) -> &'a [String];
     fn name<'a>(&'a self) -> &'a str;
@@ -73,19 +71,11 @@ impl<Axis, Layer> NeuraGraphNode<Axis, Layer> {
 
     pub fn as_boxed<Data: Clone>(self) -> Box<dyn NeuraGraphNodePartial<Data>>
     where
-        Axis: NeuraSplitInputs<Data>
+        Axis: NeuraAxis<Data>,
-            + NeuraCombineInputs<NeuraShape, Combined = Result<NeuraShape, NeuraAxisErr>>
-            + Clone
-            + Debug
-            + 'static,
         Layer: NeuraPartialLayer + Clone + Debug + 'static,
-        Layer::Constructed:
+        Layer::Constructed: NeuraLayer<Axis::Combined, Output = Data>,
-            NeuraLayer<<Axis as NeuraCombineInputs<Data>>::Combined, Output = Data>,
         Layer::Err: Debug,
-        <Layer::Constructed as NeuraLayer<
+        <Layer::Constructed as NeuraLayer<Axis::Combined>>::IntermediaryRepr: 'static,
-            <Axis as NeuraCombineInputs<Data>>::Combined,
-        >>::IntermediaryRepr: 'static,
-        <Axis as NeuraCombineInputs<Data>>::Combined: 'static,
     {
         Box::new(self)
     }
@@ -95,9 +85,8 @@ impl<Axis, Layer> NeuraGraphNode<Axis, Layer> {
         intermediary: &'a dyn Any,
     ) -> &'a Intermediary<Axis::Combined, Layer>
     where
-        Axis: NeuraCombineInputs<Data>,
+        Axis: NeuraAxis<Data>,
         Layer: NeuraLayer<Axis::Combined>,
-        Axis::Combined: 'static,
     {
         intermediary
             .downcast_ref::<Intermediary<Axis::Combined, Layer>>()
@@ -113,23 +102,16 @@ where
     layer_intermediary: Layer::IntermediaryRepr,
 }
 
-impl<
+impl<Data: Clone, Axis: NeuraAxis<Data>, Layer: NeuraLayer<Axis::Combined, Output = Data>>
-        Data: Clone,
+    NeuraGraphNodeEval<Data> for NeuraGraphNode<Axis, Layer>
-        Axis: NeuraSplitInputs<Data> + Clone + Debug,
-        Layer: NeuraLayer<<Axis as NeuraCombineInputs<Data>>::Combined, Output = Data>,
-    > NeuraGraphNodeEval<Data> for NeuraGraphNode<Axis, Layer>
-where
-    Layer::IntermediaryRepr: 'static,
-    Axis::Combined: 'static,
 {
     fn eval<'a>(&'a self, inputs: &[Data]) -> Data {
-        // TODO: use to_vec_in?
+        let combined = self.axis.combine(inputs);
-        let combined = self.axis.combine(inputs.to_vec());
         self.layer.eval(&combined)
     }
 
     fn eval_training<'a>(&self, inputs: &[Data]) -> (Data, Box<dyn Any>) {
-        let combined = self.axis.combine(inputs.to_vec());
+        let combined = self.axis.combine(inputs);
         let (result, layer_intermediary) = self.layer.eval_training(&combined);
 
         let intermediary: Intermediary<Axis::Combined, Layer> = Intermediary {
@@ -177,20 +159,11 @@ where
     }
 }
 
-impl<
+impl<Data: Clone, Axis: NeuraAxis<Data>, Layer: NeuraPartialLayer + Clone + Debug>
-        Data: Clone,
+    NeuraGraphNodePartial<Data> for NeuraGraphNode<Axis, Layer>
-        Axis: NeuraSplitInputs<Data>
-            + NeuraCombineInputs<NeuraShape, Combined = Result<NeuraShape, NeuraAxisErr>>
-            + Clone
-            + Debug
-            + 'static,
-        Layer: NeuraPartialLayer + Clone + Debug,
-    > NeuraGraphNodePartial<Data> for NeuraGraphNode<Axis, Layer>
 where
-    Layer::Constructed: NeuraLayer<<Axis as NeuraCombineInputs<Data>>::Combined, Output = Data>,
+    Layer::Constructed: NeuraLayer<Axis::Combined, Output = Data>,
     Layer::Err: Debug,
-    <Layer::Constructed as NeuraLayer<<Axis as NeuraCombineInputs<Data>>::Combined>>::IntermediaryRepr: 'static,
-    <Axis as NeuraCombineInputs<Data>>::Combined: 'static,
 {
     fn inputs<'a>(&'a self) -> &'a [String] {
         &self.inputs
@@ -206,7 +179,7 @@ where
     ) -> Result<(Box<dyn NeuraGraphNodeEval<Data>>, NeuraShape), String> {
         let combined = self
             .axis
-            .combine(input_shapes.clone())
+            .shape(&input_shapes)
             .map_err(|err| format!("{:?}", err))?;
 
         let constructed_layer = self
@@ -222,7 +195,7 @@ where
                 axis: self.axis.clone(),
                 layer: constructed_layer,
                 name: self.name.clone(),
-                input_shapes: Some(input_shapes)
+                input_shapes: Some(input_shapes),
             }),
             output_shape,
         ))

src/network/residual/construct.rs

@@ -1,4 +1,4 @@
-use crate::err::*;
+use crate::{axis::NeuraAxisBase, err::*};
 
 use super::*;
 use NeuraResidualConstructErr::*;
@@ -15,13 +15,12 @@ pub trait NeuraResidualConstruct {
     ) -> Result<Self::Constructed, Self::Err>;
 }
 
-impl<Layer: NeuraPartialLayer, ChildNetwork: NeuraResidualConstruct, Axis> NeuraResidualConstruct
+impl<Layer: NeuraPartialLayer, ChildNetwork: NeuraResidualConstruct, Axis: NeuraAxisBase>
-    for NeuraResidualNode<Layer, ChildNetwork, Axis>
+    NeuraResidualConstruct for NeuraResidualNode<Layer, ChildNetwork, Axis>
-where
-    Axis: NeuraCombineInputs<NeuraShape, Combined = Result<NeuraShape, NeuraAxisErr>>,
 {
     type Constructed = NeuraResidualNode<Layer::Constructed, ChildNetwork::Constructed, Axis>;
-    type Err = NeuraRecursiveErr<NeuraResidualConstructErr<Layer::Err>, ChildNetwork::Err>;
+    type Err =
+        NeuraRecursiveErr<NeuraResidualConstructErr<Layer::Err, Axis::Err>, ChildNetwork::Err>;
 
     fn construct_residual(
         self,
@@ -36,7 +35,7 @@ where
 
         let layer_input_shape = self
             .axis
-            .combine(input_shapes)
+            .shape(&input_shapes.iter().map(|shape| **shape).collect::<Vec<_>>())
             .map_err(|e| NeuraRecursiveErr::Current(AxisErr(e)))?;
 
         let layer = self

src/network/residual/last.rs

@@ -31,7 +31,7 @@ impl Default for NeuraResidualLast {
 
 impl NeuraResidualConstruct for NeuraResidualLast {
     type Constructed = NeuraResidualLast;
-    type Err = NeuraRecursiveErr<NeuraResidualConstructErr<()>, ()>;
+    type Err = NeuraRecursiveErr<NeuraResidualConstructErr<(), NeuraAxisErr>, ()>;
 
     fn construct_residual(
         self,

src/network/residual/mod.rs

@@ -8,9 +8,6 @@ pub use wrapper::*;
 mod input;
 pub use input::*;
 
-mod axis;
-pub use axis::*;
-
 mod construct;
 pub use construct::NeuraResidualConstruct;
 

src/network/residual/node.rs

@@ -1,6 +1,6 @@
 use std::borrow::Cow;
 
-use crate::network::*;
+use crate::{axis::*, network::*};
 
 use super::*;
 
@@ -68,12 +68,12 @@ impl<Layer, ChildNetwork, Axis> NeuraResidualNode<Layer, ChildNetwork, Axis> {
         input: &NeuraResidualInput<Data>,
     ) -> (Axis::Combined, NeuraResidualInput<Data>)
     where
-        Axis: NeuraCombineInputs<Data>,
+        Axis: NeuraAxis<Data>,
         Layer: NeuraLayer<Axis::Combined>,
     {
         let (inputs, rest) = input.shift();
 
-        let layer_input = self.axis.combine(inputs);
+        let layer_input = self.axis.combine(&inputs);
 
         (layer_input, rest)
     }
@@ -94,9 +94,9 @@ impl<Layer, ChildNetwork, Axis> NeuraResidualNode<Layer, ChildNetwork, Axis> {
 
     pub(crate) fn map_input_owned<Data>(&self, input: &NeuraResidualInput<Data>) -> Axis::Combined
     where
-        Axis: NeuraCombineInputs<Data>,
+        Axis: NeuraAxis<Data>,
     {
-        self.axis.combine(input.shift().0)
+        self.axis.combine(&input.shift().0)
     }
 }
 
@@ -148,7 +148,7 @@ impl<
 impl<Data: Clone + 'static, Layer, ChildNetwork, Axis: Clone + std::fmt::Debug + 'static>
     NeuraLayer<NeuraResidualInput<Data>> for NeuraResidualNode<Layer, ChildNetwork, Axis>
 where
-    Axis: NeuraCombineInputs<Data>,
+    Axis: NeuraAxis<Data>,
     Layer: NeuraLayer<Axis::Combined, Output = Data>,
     ChildNetwork: NeuraLayer<NeuraResidualInput<Data>>,
 {
@@ -239,7 +239,7 @@ impl<
 
 impl<
         Data: Clone + std::fmt::Debug,
-        Axis: NeuraCombineInputs<Data> + NeuraSplitInputs<Data>,
+        Axis: NeuraAxis<Data>,
         Layer: NeuraLayer<Axis::Combined, Output = Data> + std::fmt::Debug,
         ChildNetwork,
     > NeuraNetwork<NeuraResidualInput<Data>> for NeuraResidualNode<Layer, ChildNetwork, Axis>