diff --git a/examples/Basics/main.jl b/examples/Basics/main.jl
index dc13e2d92..1dd56eac5 100644
--- a/examples/Basics/main.jl
+++ b/examples/Basics/main.jl
@@ -116,10 +116,14 @@ W * x
 # the `cu` function (or the `gpu` function exported by `Lux``), and it supports all of the
 # above operations with the same syntax.
 
-using LuxCUDA
+using LuxCUDA, LuxAMDGPU
+
 if LuxCUDA.functional()
     x_cu = cu(rand(5, 3))
     @show x_cu
+elseif LuxAMDGPU.functional() # Similarly, for AMDGPU
+    x_amd = roc(rand(5, 3))
+    @show x_amd
 end
 
 # ## (Im)mutability
diff --git a/examples/DDIM/generate.jl b/examples/DDIM/generate.jl
index 83002d077..bf8209394 100644
--- a/examples/DDIM/generate.jl
+++ b/examples/DDIM/generate.jl
@@ -41,7 +41,7 @@ end
 
     mkpath(output_dir)
 
-    if CUDA.functional()
+    if LuxCUDA.functional() || LuxAMDGPU.functional()
         println("GPU is available.")
     else
         println("GPU is not available.")
diff --git a/examples/GravitationalWaveForm/main.jl b/examples/GravitationalWaveForm/main.jl
index 030aae2ae..d8c5169ef 100644
--- a/examples/GravitationalWaveForm/main.jl
+++ b/examples/GravitationalWaveForm/main.jl
@@ -17,6 +17,7 @@ close(pkg_io) #hide
 using Lux, ComponentArrays, LineSearches, LuxAMDGPU, LuxCUDA, OrdinaryDiffEq, Optimization,
       OptimizationOptimJL, Random, SciMLSensitivity
 using CairoMakie, MakiePublication
+
 CUDA.allowscalar(false)
 
 # ## Define some Utility Functions
diff --git a/examples/HyperNet/main.jl b/examples/HyperNet/main.jl
index 9d34b76f9..22aff09f7 100644
--- a/examples/HyperNet/main.jl
+++ b/examples/HyperNet/main.jl
@@ -11,6 +11,7 @@ Pkg.precompile(; io=pkg_io) #hide
 close(pkg_io) #hide
 using Lux, ComponentArrays, LuxAMDGPU, LuxCUDA, MLDatasets, MLUtils, OneHotArrays,
       Optimisers, Random, Setfield, Statistics, Zygote
+
 CUDA.allowscalar(false)
 
 # ## Loading Datasets
diff --git a/examples/ImageNet/Project.toml b/examples/ImageNet/Project.toml
index 2494c143b..9221bbf01 100644
--- a/examples/ImageNet/Project.toml
+++ b/examples/ImageNet/Project.toml
@@ -4,6 +4,7 @@ Boltz = "4544d5e4-abc5-4dea-817f-29e4c205d9c8"
 Configurations = "5218b696-f38b-4ac9-8b61-a12ec717816d"
 Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
 FLoops = "cc61a311-1640-44b5-9fba-1b764f453329"
+Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
 FluxMPI = "acf642fa-ee0e-513a-b9d5-bcd150f7ec3b"
 Formatting = "59287772-0a20-5a39-b81b-1366585eb4c0"
 Functors = "d9f16b24-f501-4c13-a1f2-28368ffc5196"
@@ -14,6 +15,7 @@ Lux = "b2108857-7c20-44ae-9111-449ecde12c47"
 LuxAMDGPU = "83120cb1-ca15-4f04-bf3b-6967d2e6b60b"
 LuxCUDA = "d0bbae9a-e099-4d5b-a835-1c6931763bda"
 MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"
+Metalhead = "dbeba491-748d-5e0e-a39e-b530a07fa0cc"
 OneHotArrays = "0b1bfda6-eb8a-41d2-88d8-f5af5cad476f"
 Optimisers = "3bd65402-5787-11e9-1adc-39752487f4e2"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
diff --git a/examples/ImageNet/main.jl b/examples/ImageNet/main.jl
index 19ad46100..25476ebe7 100644
--- a/examples/ImageNet/main.jl
+++ b/examples/ImageNet/main.jl
@@ -2,6 +2,8 @@
 
 using Augmentor                                         # Image Augmentation
 using Boltz                                             # Computer Vision Models
+import Flux                                             # Some Blotz Models need Flux
+import Metalhead                                        # Some Blotz Models need MetalHead
 using Configurations                                    # Experiment Configurations
 using LuxAMDGPU                                         # AMDGPUs <3
 using LuxCUDA                                           # NVIDIA GPUs <3
@@ -62,7 +64,8 @@ function construct(rng::AbstractRNG, cfg::ModelConfig, ecfg::ExperimentConfig)
 end
 
 function construct(cfg::OptimizerConfig)
-    if cfg.name == "adam" + opt = Adam(cfg.learning_rate)
+    if cfg.name == "adam"
+        opt = Adam(cfg.learning_rate)
     elseif cfg.name == "sgd"
         if cfg.nesterov
             opt = Nesterov(cfg.learning_rate, cfg.momentum)
diff --git a/examples/ImageNet/utils.jl b/examples/ImageNet/utils.jl
index f3c581ec5..a8363a5a9 100644
--- a/examples/ImageNet/utils.jl
+++ b/examples/ImageNet/utils.jl
@@ -1,7 +1,13 @@
 CUDA.allowscalar(false)
 
-# unsafe_free OneHotArrays
-CUDA.unsafe_free!(x::OneHotArray) = CUDA.unsafe_free!(x.indices)
+
+if LuxCUDA.functional()
+    # unsafe_free OneHotArrays
+    CUDA.unsafe_free!(x::OneHotArray) = CUDA.unsafe_free!(x.indices)
+elseif LuxAMDGPU.functional()
+    # unsafe_free OneHotArrays
+    AMDGPU.unsafe_free!(x::OneHotArray) = AMDGPU.unsafe_free!(x.indices)
+end
 
 # Loss Function
 logitcrossentropyloss(ŷ, y) = mean(-sum(y .* logsoftmax(ŷ; dims=1); dims=1))
diff --git a/examples/NeuralODE/main.jl b/examples/NeuralODE/main.jl
index 662b9d897..37a419c83 100644
--- a/examples/NeuralODE/main.jl
+++ b/examples/NeuralODE/main.jl
@@ -18,6 +18,7 @@ using Lux, ComponentArrays, SciMLSensitivity, LuxAMDGPU, LuxCUDA, Optimisers,
       OrdinaryDiffEq, Random, Statistics, Zygote, OneHotArrays, InteractiveUtils
 import MLDatasets: MNIST
 import MLUtils: DataLoader, splitobs
+
 CUDA.allowscalar(false)
 
 # ## Loading MNIST
diff --git a/examples/SimpleRNN/main.jl b/examples/SimpleRNN/main.jl
index cfff2185e..16d8681f6 100644
--- a/examples/SimpleRNN/main.jl
+++ b/examples/SimpleRNN/main.jl
@@ -84,7 +84,7 @@ function (s::SpiralClassifier)(
     ## See that the parameters and states are automatically populated into a field called
     ## `lstm_cell` We use `eachslice` to get the elements in the sequence without copying,
     ## and `Iterators.peel` to split out the first element for LSTM initialization.
-    x_init, x_rest = Iterators.peel(eachslice(x; dims=2))
+    x_init, x_rest = Iterators.peel(Lux._eachslice(x, Val(2)))
     (y, carry), st_lstm = s.lstm_cell(x_init, ps.lstm_cell, st.lstm_cell)
     ## Now that we have the hidden state and memory in `carry` we will pass the input and
     ## `carry` jointly
@@ -119,7 +119,7 @@ function compute_loss(x, y, model, ps, st)
     return binarycrossentropy(y_pred, y), y_pred, st
 end
 
-matches(y_pred, y_true) = sum((y_pred .> 0.5) .== y_true)
+matches(y_pred, y_true) = sum((y_pred .> 0.5f0) .== y_true)
 accuracy(y_pred, y_true) = matches(y_pred, y_true) / length(y_pred)
 
 # Finally lets create an optimiser given the model parameters.