diff --git a/src/layers/rotary.jl b/src/layers/rotary.jl
index f7f8d08d13..8c90bf68a0 100644
--- a/src/layers/rotary.jl
+++ b/src/layers/rotary.jl
@@ -1,19 +1,19 @@
 """
     Rotary Position Embeddings (RoPE)
 
-This is a port of the RoPE implementation from NeuralAttentionlib.jl, which is an implementation of
+This is a port and simplified code of the RoPE implementation from NeuralAttentionlib.jl, which is an implementation of
 the Rotary Position Embeddings (RoPE) described in the RoFormer paper.
 
 Original sources:
 - Paper: "RoFormer: Enhanced Transformer with Rotary Position Embedding"
   Authors: Jianlin Su, Yu Lu, Shengfeng Pan, Ahmed Murtadha, Bo Wen
   URL: https://arxiv.org/abs/2104.09864
-  
+
 - Code: NeuralAttentionlib.jl
   Author: chengchingwen
   Repository: https://github.com/chengchingwen/NeuralAttentionlib.jl
 
-RoPE encodes absolute positional information with a rotation matrix that naturally 
+RoPE encodes absolute positional information with a rotation matrix that naturally
 incorporates explicit relative position dependency in self-attention formulation.
 """
 
diff --git a/test/layers/rotary.jl b/test/layers/rotary.jl
new file mode 100644
index 0000000000..8fcb98105a
--- /dev/null
+++ b/test/layers/rotary.jl
@@ -0,0 +1,32 @@
+using Flux: with_rotary_position_embedding
+
+@testset "Rotary Position Embedding Tests" begin
+    Random.seed!(123)
+    test_sizes = [(2,2), (4,6), (8,10)]
+
+    for (n, d) in test_sizes
+        x = randn(n, d)
+        test_gradients(
+            with_rotary_position_embedding,
+            x;
+            rtol=1e-4,
+            atol=1e-4,
+            test_gpu=false,
+            compare_finite_diff=true,
+            loss=(f, x) -> sum(f(x))
+        )
+    end
+
+    # Edge cases
+    test_gradients(
+        with_rotary_position_embedding,
+        zeros(4, 6);
+        loss=(f, x) -> sum(f(x))
+    )
+
+    test_gradients(
+        with_rotary_position_embedding,
+        ones(4, 6);
+        loss=(f, x) -> sum(f(x))
+    )
+end
diff --git a/test/runtests.jl b/test/runtests.jl
index 6f5a2e7d84..983ee67eab 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -55,6 +55,7 @@ Random.seed!(0)
       include("layers/upsample.jl")
       include("layers/show.jl")
       include("layers/macro.jl")
+      include("layers/rotary.jl")
     end
 
     @testset "outputsize" begin