From e2f060fa816ba544db91251714eace59e5fa9f9c Mon Sep 17 00:00:00 2001
From: vincent <captainvincenttw@gmail.com>
Date: Wed, 15 May 2024 21:10:40 +0800
Subject: [PATCH] [Squeezenet] Upgrade to burn 0.13.0 (#33)

* [Squeezenet] Upgrade to burn 0.13.0

* Upgrade to burn 0.13.0
* Add feature flag to disable copy the weights file

Signed-off-by: vincent <vincent@secondstate.io>

* [CI] Update the minimum Rust version to 1.75

Signed-off-by: vincent <vincent@secondstate.io>

---------

Signed-off-by: vincent <vincent@secondstate.io>
---
 .github/workflows/rust.yml              |  2 +-
 squeezenet-burn/Cargo.toml              | 11 +++++++----
 squeezenet-burn/build.rs                |  2 +-
 squeezenet-burn/examples/classify.rs    | 10 ++++++----
 squeezenet-burn/src/model/normalizer.rs | 12 +++---------
 5 files changed, 18 insertions(+), 19 deletions(-)

diff --git a/.github/workflows/rust.yml b/.github/workflows/rust.yml
index c01d36b..10facb0 100644
--- a/.github/workflows/rust.yml
+++ b/.github/workflows/rust.yml
@@ -15,7 +15,7 @@ jobs:
 
     strategy:
       matrix:
-        rust: [stable, 1.74.0]
+        rust: [stable, 1.75.0]
 
     steps:
       - uses: actions/checkout@v4
diff --git a/squeezenet-burn/Cargo.toml b/squeezenet-burn/Cargo.toml
index 40e3977..8721ecf 100644
--- a/squeezenet-burn/Cargo.toml
+++ b/squeezenet-burn/Cargo.toml
@@ -6,7 +6,7 @@ version = "0.1.0"
 edition = "2021"
 
 [features]
-default = ["weights_file"]
+default = ["weights_file", "weights_file_dump"]
 
 # Enables Half precision (f16) support
 weights_f16 = []
@@ -17,11 +17,14 @@ weights_embedded = []
 # Use weights from a file
 weights_file = ["burn/default"]
 
+# Copy weights file to specif folder
+weights_file_dump = []
+
 
 [dependencies]
 
 # Note: default-features = false is needed to disable std
-burn = { version = "0.11.1", default-features = false }
+burn = { version = "0.13.0", default-features = false }
 
 # Used to load weights from a file
 serde = { version = "1.0.183", default-features = false, features = [
@@ -31,9 +34,9 @@ serde = { version = "1.0.183", default-features = false, features = [
 
 [dev-dependencies]
 # Used by the classify example
-burn = { version = "0.11.1", features = ["ndarray"]  }
+burn = { version = "0.13.0", features = ["ndarray"]  }
 image = { version = "0.24.7", features = ["png", "jpeg"] }
 
 [build-dependencies]
 # Used to generate code from ONNX model
-burn-import = { version = "0.11.1", package = "burn-import" }
+burn-import = { version = "0.13.2", package = "burn-import" }
diff --git a/squeezenet-burn/build.rs b/squeezenet-burn/build.rs
index bb8b3f8..b9dffd6 100644
--- a/squeezenet-burn/build.rs
+++ b/squeezenet-burn/build.rs
@@ -47,7 +47,7 @@ fn main() {
         .run_from_script();
 
     // Copy the weights next to the executable.
-    if cfg!(feature = "weights_file") {
+    if cfg!(feature = "weights_file") && cfg!(feature = "weights_file_dump") {
         copy_weights_next_to_executable();
     }
 
diff --git a/squeezenet-burn/examples/classify.rs b/squeezenet-burn/examples/classify.rs
index 9900e9c..b367c4e 100644
--- a/squeezenet-burn/examples/classify.rs
+++ b/squeezenet-burn/examples/classify.rs
@@ -42,11 +42,13 @@ fn main() {
         }
     }
 
-    // Create a tensor from the array
-    let image_input = Tensor::<Backend, 3>::from_data(img_array).reshape([1, 3, HEIGHT, WIDTH]);
+    let device = Default::default();
 
+    // Create a tensor from the array
+    let image_input =
+        Tensor::<Backend, 3>::from_data(img_array, &device).reshape([1, 3, HEIGHT, WIDTH]);
     // Normalize the image
-    let normalizer = Normalizer::new();
+    let normalizer = Normalizer::new(&device);
     let normalized_image = normalizer.normalize(image_input);
 
     // Create the model
@@ -59,7 +61,7 @@ fn main() {
         .join(RECORD_FILE);
 
     #[cfg(feature = "weights_file")]
-    let model = Model::<Backend>::from_file(weights_file.to_str().unwrap());
+    let model = Model::<Backend>::from_file(weights_file.to_str().unwrap(), &device);
 
     #[cfg(feature = "weights_embedded")]
     // Load model from embedded weights
diff --git a/squeezenet-burn/src/model/normalizer.rs b/squeezenet-burn/src/model/normalizer.rs
index 7e1e601..f14439e 100644
--- a/squeezenet-burn/src/model/normalizer.rs
+++ b/squeezenet-burn/src/model/normalizer.rs
@@ -13,9 +13,9 @@ pub struct Normalizer<B: Backend> {
 
 impl<B: Backend> Normalizer<B> {
     /// Creates a new normalizer.
-    pub fn new() -> Self {
-        let mean = Tensor::from_floats(MEAN).reshape([1, 3, 1, 1]);
-        let std = Tensor::from_floats(STD).reshape([1, 3, 1, 1]);
+    pub fn new(device: &B::Device) -> Self {
+        let mean = Tensor::from_floats(MEAN, device).reshape([1, 3, 1, 1]);
+        let std = Tensor::from_floats(STD, device).reshape([1, 3, 1, 1]);
         Self { mean, std }
     }
 
@@ -30,9 +30,3 @@ impl<B: Backend> Normalizer<B> {
         (input - self.mean.clone()) / self.std.clone()
     }
 }
-
-impl<B: Backend> Default for Normalizer<B> {
-    fn default() -> Self {
-        Self::new()
-    }
-}