Update to Edition 2024 and bump deps

Cargo.toml

@@ -1,22 +1,22 @@
 [package]
+rust-version = "1.85.0"
 name = "tensorflow-layers"
-version = "0.4.0"
+version = "0.5.0"
 authors = ["Crowdstrike DSCI <dsci-oss@crowdstrike.com>"]
-edition = "2021"
-rust = "1.60.0"
+edition = "2024"
 description = "Pure Rust implementation of layers used in Tensorflow models"
 license = "MIT"
 include = ["Cargo.toml", "README.md", "benches/*", "src/*"]
 
 [dependencies]
-ndarray = { version = "0.15.5", features = ["serde-1"] }
-ndarray-rand = "0.14.0"
-num-traits = "0.2.14"
-serde = { version = "1.0.188", features = ["derive"] }
+ndarray = { version = "0.17.2", features = ["serde"] }
+ndarray-rand = "0.16.0"
+num-traits = "0.2.19"
+serde = { version = "1.0.228", features = ["derive"] }
 
 [dev-dependencies]
-criterion = "0.5.1"
-lazy_static = "1.4.0"
+criterion = "0.8.1"
+lazy_static = "1.5.0"
 
 [[bench]]
 name = "benchmarks"

benches/benchmarks.rs

@@ -4,7 +4,7 @@
 #[macro_use]
 extern crate lazy_static;
 
-use criterion::{black_box, criterion_group, criterion_main, Criterion};
+use criterion::{Criterion, black_box, criterion_group, criterion_main};
 use ndarray::*;
 use tensorflow_layers::*;
 

src/activations.rs

@@ -131,7 +131,7 @@ impl Activation {
 #[allow(clippy::unreadable_literal)]
 mod tests {
     use super::*;
-    use ndarray::{array, Array1, Array2, Array3};
+    use ndarray::{Array1, Array2, Array3, array};
 
     #[test]
     fn test_relu_1d() {

src/average_pooling.rs

@@ -101,7 +101,7 @@ impl AveragePooling1DLayer {
 #[allow(clippy::unreadable_literal)]
 mod tests {
     use super::*;
-    use ndarray::{array, Array1, Array2, Array3};
+    use ndarray::{Array1, Array2, Array3, array};
 
     #[test]
     fn test_averagepooling1d() {
@@ -192,12 +192,20 @@ mod tests {
                 [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
             ],
             [
-                [0.0, 5.73501, 6.6221056, 5.9534774, 2.8128617, 1.6962836, 0.0],
-                [0.0, -0.810096, 0.7763562, 2.1465158, 4.916585, 5.012854, 0.0]
+                [
+                    0.0, 5.73501, 6.6221056, 5.9534774, 2.8128617, 1.6962836, 0.0
+                ],
+                [
+                    0.0, -0.810096, 0.7763562, 2.1465158, 4.916585, 5.012854, 0.0
+                ]
             ],
             [
-                [0.0, 3.8437622, 0.8588956, 4.1046715, 3.6405258, 2.4107575, 0.0],
-                [0.0, 1.9883004, 3.3759031, 1.6852531, 0.7157619, 1.3628602, 0.0]
+                [
+                    0.0, 3.8437622, 0.8588956, 4.1046715, 3.6405258, 2.4107575, 0.0
+                ],
+                [
+                    0.0, 1.9883004, 3.3759031, 1.6852531, 0.7157619, 1.3628602, 0.0
+                ]
             ]
         ];
 

src/batch_normalization.rs

@@ -76,7 +76,7 @@ impl BatchNormalization {
 #[allow(clippy::unreadable_literal)]
 mod tests {
     use super::*;
-    use ndarray::{arr2, arr3, Array2, Array3};
+    use ndarray::{Array2, Array3, arr2, arr3};
 
     #[test]
     fn test_batchnormalization_1d() {

src/conv1d.rs

@@ -46,7 +46,10 @@ impl Conv1DLayer {
                 weights.raw_dim()[0],
                 weights.raw_dim()[1] * weights.raw_dim()[2],
             ];
-            weights.into_shape(output_shape).unwrap().reversed_axes()
+            weights
+                .into_shape_with_order(output_shape)
+                .unwrap()
+                .reversed_axes()
         };
 
         Conv1DLayer {
@@ -146,7 +149,7 @@ impl Conv1DLayer {
         // is reconstructed back from 2D to 3D.
         let mut result = data_intermediate
             .dot(&self.weights)
-            .into_shape(out_shape)
+            .into_shape_with_order(out_shape)
             .unwrap();
 
         // Apply the bias
@@ -162,7 +165,7 @@ impl Conv1DLayer {
 #[allow(clippy::unreadable_literal)]
 mod tests {
     use super::*;
-    use ndarray::{arr3, Array, Array1, Array3};
+    use ndarray::{Array, Array1, Array3, arr3};
 
     #[test]
     fn test_conv1d_simple() {

src/dense.rs

@@ -78,7 +78,7 @@ impl DenseLayer {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use ndarray::{array, Array, Array1, Array2, Array3};
+    use ndarray::{Array, Array1, Array2, Array3, array};
 
     #[test]
     fn test_dense_2d_simple() {
@@ -121,9 +121,15 @@ mod tests {
 
         let result = dense_layer.apply2d(&data);
         let expected = array![
-            [3587.0, 3654.0, 3721.0, 3788.0, 3855.0, 3922.0, 3989.0, 4056.0],
-            [8548.0, 8736.0, 8924.0, 9112.0, 9300.0, 9488.0, 9676.0, 9864.0],
-            [13509.0, 13818.0, 14127.0, 14436.0, 14745.0, 15054.0, 15363.0, 15672.0]
+            [
+                3587.0, 3654.0, 3721.0, 3788.0, 3855.0, 3922.0, 3989.0, 4056.0
+            ],
+            [
+                8548.0, 8736.0, 8924.0, 9112.0, 9300.0, 9488.0, 9676.0, 9864.0
+            ],
+            [
+                13509.0, 13818.0, 14127.0, 14436.0, 14745.0, 15054.0, 15363.0, 15672.0
+            ]
         ];
 
         assert_eq!(expected, result);
@@ -141,14 +147,26 @@ mod tests {
         let result = dense_layer.apply3d(&data);
         let expected = array![
             [
-                [3587.0, 3654.0, 3721.0, 3788.0, 3855.0, 3922.0, 3989.0, 4056.0],
-                [8548.0, 8736.0, 8924.0, 9112.0, 9300.0, 9488.0, 9676.0, 9864.0],
-                [13509.0, 13818.0, 14127.0, 14436.0, 14745.0, 15054.0, 15363.0, 15672.0]
+                [
+                    3587.0, 3654.0, 3721.0, 3788.0, 3855.0, 3922.0, 3989.0, 4056.0
+                ],
+                [
+                    8548.0, 8736.0, 8924.0, 9112.0, 9300.0, 9488.0, 9676.0, 9864.0
+                ],
+                [
+                    13509.0, 13818.0, 14127.0, 14436.0, 14745.0, 15054.0, 15363.0, 15672.0
+                ]
             ],
             [
-                [18470.0, 18900.0, 19330.0, 19760.0, 20190.0, 20620.0, 21050.0, 21480.0],
-                [23431.0, 23982.0, 24533.0, 25084.0, 25635.0, 26186.0, 26737.0, 27288.0],
-                [28392.0, 29064.0, 29736.0, 30408.0, 31080.0, 31752.0, 32424.0, 33096.0]
+                [
+                    18470.0, 18900.0, 19330.0, 19760.0, 20190.0, 20620.0, 21050.0, 21480.0
+                ],
+                [
+                    23431.0, 23982.0, 24533.0, 25084.0, 25635.0, 26186.0, 26737.0, 27288.0
+                ],
+                [
+                    28392.0, 29064.0, 29736.0, 30408.0, 31080.0, 31752.0, 32424.0, 33096.0
+                ]
             ]
         ];
 

src/dropout.rs

@@ -1,5 +1,5 @@
 use ndarray::{Array2, Zip};
-use ndarray_rand::{rand, rand_distr::Bernoulli, RandomExt};
+use ndarray_rand::{RandomExt, rand, rand_distr::Bernoulli};
 use serde::{Deserialize, Serialize};
 
 /// The implementation of Dropout layer
@@ -36,7 +36,7 @@ impl Dropout {
         }
 
         let mut res: Array2<f32> = data.clone();
-        let mut rng = rand::thread_rng();
+        let mut rng = rand::rng();
 
         let mask2 = Array2::random_using(
             data.dim(),

src/embedding.rs

@@ -54,14 +54,14 @@ impl EmbeddingLayer {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use ndarray::{arr3, array, Array, Array2, Array3, Array4};
+    use ndarray::{Array, Array2, Array3, Array4, arr3, array};
 
     #[test]
     fn test_embedding_2d() {
         let data: Array2<usize> = array![[1, 2, 3], [4, 5, 6], [7, 8, 9], [1, 4, 7]];
 
         let weights: Array2<f32> = Array::linspace(1., 1000., 1000)
-            .into_shape([100, 10])
+            .into_shape_with_order([100, 10])
             .unwrap();
         let embedding_layer = EmbeddingLayer::new(weights);
 
@@ -96,7 +96,7 @@ mod tests {
         let data: Array2<u16> = array![[1, 2, 3], [4, 5, 6], [7, 8, 9], [1, 4, 7]];
 
         let weights: Array2<f32> = Array::linspace(1., 1000., 1000)
-            .into_shape([100, 10])
+            .into_shape_with_order([100, 10])
             .unwrap();
         let embedding_layer = EmbeddingLayer::new(weights);
 
@@ -131,7 +131,7 @@ mod tests {
         let data: Array3<usize> = array![[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [1, 4, 7]]];
 
         let weights: Array2<f32> = Array::linspace(1., 1000., 1000)
-            .into_shape([100, 10])
+            .into_shape_with_order([100, 10])
             .unwrap();
         let embedding_layer = EmbeddingLayer::new(weights);
 
@@ -160,7 +160,7 @@ mod tests {
         let data: Array2<usize> = array![[1, 2, 3], [4, 5, 100], [7, 8, 9], [1, 4, 7]];
 
         let weights: Array2<f32> = Array::linspace(1., 1000., 1000)
-            .into_shape([100, 10])
+            .into_shape_with_order([100, 10])
             .unwrap();
         let embedding_layer = EmbeddingLayer::new(weights);
 
@@ -173,7 +173,7 @@ mod tests {
         let data: Array3<usize> = array![[[1, 2, 3], [4, 5, 100]], [[7, 8, 9], [1, 4, 7]]];
 
         let weights: Array2<f32> = Array::linspace(1., 1000., 1000)
-            .into_shape([100, 10])
+            .into_shape_with_order([100, 10])
             .unwrap();
         let embedding_layer = EmbeddingLayer::new(weights);
 

src/max_pooling.rs

@@ -95,7 +95,7 @@ impl MaxPooling1DLayer {
 #[allow(clippy::unreadable_literal)]
 mod tests {
     use super::*;
-    use ndarray::{array, Array1, Array2, Array3};
+    use ndarray::{Array1, Array2, Array3, array};
 
     #[test]
     fn test_maxpooling1d() {
@@ -200,8 +200,12 @@ mod tests {
                 ]
             ],
             [
-                [0.0, 5.975927, 3.1096065, 5.774783, 4.3322973, 3.3641431, 0.0],
-                [0.0, 4.976529, 5.649868, 2.2129567, 1.3772863, 4.3095136, 0.0]
+                [
+                    0.0, 5.975927, 3.1096065, 5.774783, 4.3322973, 3.3641431, 0.0
+                ],
+                [
+                    0.0, 4.976529, 5.649868, 2.2129567, 1.3772863, 4.3095136, 0.0
+                ]
             ]
         ];