diff --git a/Torchsharp-llama/ITokenizer.cs b/Torchsharp-llama/ITokenizer.cs
index f027872..3f78d40 100644
--- a/Torchsharp-llama/ITokenizer.cs
+++ b/Torchsharp-llama/ITokenizer.cs
@@ -125,8 +125,6 @@ public int[] Encode(string input, bool bos, bool eos)
             tokens = tokens.Concat(new int[] { this.EosId }).ToArray();
         }
 
-        Console.WriteLine($"tokens: {string.Join(",", tokens)}");
-
         return tokens;
     }
 }
diff --git a/Torchsharp-llama/Program.cs b/Torchsharp-llama/Program.cs
index 162f028..997cfc8 100644
--- a/Torchsharp-llama/Program.cs
+++ b/Torchsharp-llama/Program.cs
@@ -6,30 +6,19 @@
 var vocabPath = @"vocab.json";
 var mergesPath = @"merges.txt";
 var tokenizer = new BPETokenizer(vocabPath, mergesPath);
-var checkpointDirectory = "/home/xiaoyuz/llama/llama-2-7b";
-var device = "cuda";
-
-if (device == "cuda")
-{
-    torch.InitializeDeviceType(DeviceType.CUDA);
-    torch.cuda.is_available().Should().BeTrue();
-}
-
-torch.manual_seed(100);
-var model = LLaMA.Build(
-       checkPointsDirectory: checkpointDirectory,
-       tokenizer: tokenizer,
-       maxSeqLen: 128,
-       maxBatchSize: 1,
-       device: device);
 
 var prompts = new[]
 {
     "I believe the meaning of life is",
+    "Replace this text in the input field to see how llama tokenization works."
 };
-var result = model.TextCompletion(prompts, temperature: 0, echo: true, device: device);
 
-foreach (var item in result)
+foreach (var prompt in prompts)
 {
-    Console.WriteLine($"generation: {item.generation}");
+    var tokens = tokenizer.Encode(prompt, bos: true, eos: false);
+    var decoded = tokenizer.Decode(tokens);
+    Console.WriteLine($"Prompt: {prompt}");
+    Console.WriteLine($"Tokens: {string.Join(", ", tokens)}");
+    Console.WriteLine($"Decoded: {decoded}");
+    Console.WriteLine();
 }
diff --git a/Torchsharp-llama/Torchsharp-llama.csproj b/Torchsharp-llama/Torchsharp-llama.csproj
index 4d15461..3805f6b 100644
--- a/Torchsharp-llama/Torchsharp-llama.csproj
+++ b/Torchsharp-llama/Torchsharp-llama.csproj
@@ -12,7 +12,7 @@
   <ItemGroup>
     <PackageReference Include="FluentAssertions" Version="7.0.0-alpha.3" />
     <PackageReference Include="Microsoft.ML.Tokenizers" Version="0.21.1" />
-    <PackageReference Include="TorchSharp-cuda-linux" Version="0.100.7" />
+    <PackageReference Include="TorchSharp-cuda-windows" Version="0.100.7" />
     <!-- <PackageReference Include="TorchSharp-cpu" Version="0.100.7" /> -->
     <!-- <PackageReference Include="libtorch-cuda-12.1-linux-x64" Version="2.1.0.1" /> -->
   </ItemGroup>