.Net: Bump ONNX to 0.5.2 (#9644)

### Motivation and Context - The Latest 0.5.0 package also requires the caller to handle the resources with the `OgaHandler` instance, when a service is instantiated this resource needs to be also present and exposed together with the service. Otherwise a message will be sent to the console and the application will crash before finishing. - Resolves #9628 --------- Co-authored-by: westey <[email protected]>
microsoft · Dec 2, 2024 · 4674281 · 4674281
1 parent d9fd8ff
commit 4674281
Show file tree

Hide file tree

Showing 8 changed files with 88 additions and 43 deletions.
diff --git a/dotnet/Directory.Packages.props b/dotnet/Directory.Packages.props
@@ -43,7 +43,7 @@
     <PackageVersion Include="Microsoft.CodeAnalysis.CSharp" Version="4.11.0" />
     <PackageVersion Include="Microsoft.Bcl.TimeProvider" Version="8.0.1" />
     <PackageVersion Include="Microsoft.Identity.Client" Version="4.66.2" />
-    <PackageVersion Include="Microsoft.ML.OnnxRuntime" Version="1.19.2" />
+    <PackageVersion Include="Microsoft.ML.OnnxRuntime" Version="1.20.1" />
     <PackageVersion Include="FastBertTokenizer" Version="1.0.28" />
     <PackageVersion Include="PdfPig" Version="0.1.9" />
     <PackageVersion Include="Pinecone.NET" Version="2.1.1" />
@@ -161,8 +161,8 @@
       <IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
     </PackageReference>
     <!-- OnnxRuntimeGenAI -->
-    <PackageVersion Include="Microsoft.ML.OnnxRuntimeGenAI" Version="0.4.0" />
-    <PackageVersion Include="Microsoft.ML.OnnxRuntimeGenAI.Cuda" Version="0.4.0" />
-    <PackageVersion Include="Microsoft.ML.OnnxRuntimeGenAI.DirectML" Version="0.4.0" />
+    <PackageVersion Include="Microsoft.ML.OnnxRuntimeGenAI" Version="0.5.2" />
+    <PackageVersion Include="Microsoft.ML.OnnxRuntimeGenAI.Cuda" Version="0.5.2" />
+    <PackageVersion Include="Microsoft.ML.OnnxRuntimeGenAI.DirectML" Version="0.5.2" />
   </ItemGroup>
 </Project>
diff --git a/dotnet/samples/Concepts/ChatCompletion/Onnx_ChatCompletion.cs b/dotnet/samples/Concepts/ChatCompletion/Onnx_ChatCompletion.cs
@@ -34,7 +34,7 @@ public async Task ServicePromptAsync()
 
         Console.WriteLine("======== Onnx - Chat Completion ========");
 
-        var chatService = new OnnxRuntimeGenAIChatCompletionService(
+        using var chatService = new OnnxRuntimeGenAIChatCompletionService(
             modelId: TestConfiguration.Onnx.ModelId,
             modelPath: TestConfiguration.Onnx.ModelPath);
 
@@ -105,5 +105,21 @@ public async Task ChatPromptAsync()
         reply = await kernel.InvokePromptAsync(chatPrompt.ToString());
 
         Console.WriteLine(reply);
+
+        DisposeServices(kernel);
+    }
+
+    /// <summary>
+    /// To avoid any potential memory leak all disposable services created by the kernel are disposed.
+    /// </summary>
+    /// <param name="kernel">Target kernel</param>
+    private static void DisposeServices(Kernel kernel)
+    {
+        foreach (var target in kernel
+            .GetAllServices<IChatCompletionService>()
+            .OfType<IDisposable>())
+        {
+            target.Dispose();
+        }
     }
 }
diff --git a/dotnet/samples/Concepts/ChatCompletion/Onnx_ChatCompletionStreaming.cs b/dotnet/samples/Concepts/ChatCompletion/Onnx_ChatCompletionStreaming.cs
@@ -29,18 +29,36 @@ public class Onnx_ChatCompletionStreaming(ITestOutputHelper output) : BaseTest(o
     /// </list>
     /// </remarks>
     [Fact]
-    public Task StreamChatAsync()
+    public async Task StreamChatAsync()
     {
         Assert.NotNull(TestConfiguration.Onnx.ModelId);   // dotnet user-secrets set "Onnx:ModelId" "<model-id>"
         Assert.NotNull(TestConfiguration.Onnx.ModelPath); // dotnet user-secrets set "Onnx:ModelPath" "<model-folder-path>"
 
         Console.WriteLine("======== Onnx - Chat Completion Streaming ========");
 
-        var chatService = new OnnxRuntimeGenAIChatCompletionService(
+        using var chatService = new OnnxRuntimeGenAIChatCompletionService(
             modelId: TestConfiguration.Onnx.ModelId,
             modelPath: TestConfiguration.Onnx.ModelPath);
 
-        return this.StartStreamingChatAsync(chatService);
+        Console.WriteLine("Chat content:");
+        Console.WriteLine("------------------------");
+
+        var chatHistory = new ChatHistory("You are a librarian, expert about books");
+        OutputLastMessage(chatHistory);
+
+        // First user message
+        chatHistory.AddUserMessage("Hi, I'm looking for book suggestions");
+        OutputLastMessage(chatHistory);
+
+        // First assistant message
+        await StreamMessageOutputAsync(chatService, chatHistory, AuthorRole.Assistant);
+
+        // Second user message
+        chatHistory.AddUserMessage("I love history and philosophy, I'd like to learn something new about Greece, any suggestion?");
+        OutputLastMessage(chatHistory);
+
+        // Second assistant message
+        await StreamMessageOutputAsync(chatService, chatHistory, AuthorRole.Assistant);
     }
 
     /// <summary>
@@ -86,6 +104,8 @@ public async Task StreamChatPromptAsync()
         reply = await StreamMessageOutputFromKernelAsync(kernel, chatPrompt.ToString());
 
         Console.WriteLine(reply);
+
+        DisposeServices(kernel);
     }
 
     /// <summary>
@@ -115,7 +135,7 @@ public async Task StreamTextFromChatAsync()
         Console.WriteLine("======== Stream Text from Chat Content ========");
 
         // Create chat completion service
-        var chatService = new OnnxRuntimeGenAIChatCompletionService(
+        using var chatService = new OnnxRuntimeGenAIChatCompletionService(
             modelId: TestConfiguration.Onnx.ModelId,
             modelPath: TestConfiguration.Onnx.ModelPath);
 
@@ -135,30 +155,7 @@ public async Task StreamTextFromChatAsync()
         }
     }
 
-    private async Task StartStreamingChatAsync(IChatCompletionService chatCompletionService)
-    {
-        Console.WriteLine("Chat content:");
-        Console.WriteLine("------------------------");
-
-        var chatHistory = new ChatHistory("You are a librarian, expert about books");
-        OutputLastMessage(chatHistory);
-
-        // First user message
-        chatHistory.AddUserMessage("Hi, I'm looking for book suggestions");
-        OutputLastMessage(chatHistory);
-
-        // First assistant message
-        await StreamMessageOutputAsync(chatCompletionService, chatHistory, AuthorRole.Assistant);
-
-        // Second user message
-        chatHistory.AddUserMessage("I love history and philosophy, I'd like to learn something new about Greece, any suggestion?");
-        OutputLastMessage(chatHistory);
-
-        // Second assistant message
-        await StreamMessageOutputAsync(chatCompletionService, chatHistory, AuthorRole.Assistant);
-    }
-
-    private async Task StreamMessageOutputAsync(IChatCompletionService chatCompletionService, ChatHistory chatHistory, AuthorRole authorRole)
+    private async Task StreamMessageOutputAsync(OnnxRuntimeGenAIChatCompletionService chatCompletionService, ChatHistory chatHistory, AuthorRole authorRole)
     {
         bool roleWritten = false;
         string fullMessage = string.Empty;
@@ -205,4 +202,18 @@ private async Task<string> StreamMessageOutputFromKernelAsync(Kernel kernel, str
         Console.WriteLine("\n------------------------");
         return fullMessage;
     }
+
+    /// <summary>
+    /// To avoid any potential memory leak all disposable services created by the kernel are disposed.
+    /// </summary>
+    /// <param name="kernel">Target kernel</param>
+    private static void DisposeServices(Kernel kernel)
+    {
+        foreach (var target in kernel
+            .GetAllServices<IChatCompletionService>()
+            .OfType<IDisposable>())
+        {
+            target.Dispose();
+        }
+    }
 }
diff --git a/dotnet/samples/Demos/OnnxSimpleRAG/OnnxSimpleRAG.csproj b/dotnet/samples/Demos/OnnxSimpleRAG/OnnxSimpleRAG.csproj
@@ -3,7 +3,7 @@
   <PropertyGroup>
     <OutputType>Exe</OutputType>
     <TargetFramework>net8.0</TargetFramework>
-    <NoWarn>$(NoWarn);CA2007;CS0612;VSTHRD111</NoWarn>
+    <NoWarn>$(NoWarn);CA2007;CS0612;VSTHRD111;SKEXP0070;SKEXP0050;SKEXP0001;SKEXP0020</NoWarn>
     <UserSecretsId>5ee045b0-aea3-4f08-8d31-32d1a6f8fed0</UserSecretsId>
   </PropertyGroup>
 

diff --git a/dotnet/samples/Demos/OnnxSimpleRAG/Program.cs b/dotnet/samples/Demos/OnnxSimpleRAG/Program.cs
@@ -1,17 +1,15 @@
 // Copyright (c) Microsoft. All rights reserved.
 
-#pragma warning disable SKEXP0070
-#pragma warning disable SKEXP0050
-#pragma warning disable SKEXP0001
-#pragma warning disable SKEXP0020
-
 using System;
 using System.IO;
+using System.Linq;
 using Microsoft.Extensions.Configuration;
 using Microsoft.Extensions.VectorData;
+using Microsoft.ML.OnnxRuntimeGenAI;
 using Microsoft.SemanticKernel;
 using Microsoft.SemanticKernel.ChatCompletion;
 using Microsoft.SemanticKernel.Connectors.InMemory;
+using Microsoft.SemanticKernel.Connectors.Onnx;
 using Microsoft.SemanticKernel.Data;
 using Microsoft.SemanticKernel.Embeddings;
 using Microsoft.SemanticKernel.PromptTemplates.Handlebars;
@@ -29,6 +27,10 @@
 // Path to the vocab file your ONNX BGE-MICRO-V2 model
 var embeddingVocabPath = config["Onnx:EmbeddingVocabPath"]!;
 
+// If using Onnx GenAI 0.5.0 or later, the OgaHandle class must be used to track
+// resources used by the Onnx services, before using any of the Onnx services.
+using var ogaHandle = new OgaHandle();
+
 // Load the services
 var builder = Kernel.CreateBuilder()
     .AddOnnxRuntimeGenAIChatCompletion(chatModelId, chatModelPath)
@@ -38,7 +40,7 @@
 var kernel = builder.Build();
 
 // Get the instances of the services
-var chatService = kernel.GetRequiredService<IChatCompletionService>();
+using var chatService = kernel.GetRequiredService<IChatCompletionService>() as OnnxRuntimeGenAIChatCompletionService;
 var embeddingService = kernel.GetRequiredService<ITextEmbeddingGenerationService>();
 
 // Create a vector store and a collection to store information
@@ -71,9 +73,12 @@ await collection.UpsertAsync(new()
     Console.Write("User > ");
     var question = Console.ReadLine()!;
 
+    // Clean resources and exit the demo if the user input is null or empty
     if (question is null || string.IsNullOrWhiteSpace(question))
     {
-        // Exit the demo if the user input is null or empty
+        // To avoid any potential memory leak all disposable
+        // services created by the kernel are disposed
+        DisposeServices(kernel);
         return;
     }
 
@@ -105,6 +110,19 @@ await collection.UpsertAsync(new()
     Console.WriteLine();
 }
 
+static void DisposeServices(Kernel kernel)
+{
+    foreach (var target in kernel
+        .GetAllServices<IChatCompletionService>()
+        .OfType<IDisposable>())
+    {
+        target.Dispose();
+    }
+}
+
+/// <summary>
+/// Information item to represent the embedding data stored in the memory
+/// </summary>
 internal sealed class InformationItem
 {
     [VectorStoreRecordKey]

diff --git a/dotnet/src/Connectors/Connectors.Onnx/Connectors.Onnx.csproj b/dotnet/src/Connectors/Connectors.Onnx/Connectors.Onnx.csproj
@@ -5,8 +5,8 @@
     <AssemblyName>Microsoft.SemanticKernel.Connectors.Onnx</AssemblyName>
     <RootNamespace>$(AssemblyName)</RootNamespace>
     <TargetFrameworks>net8.0;netstandard2.0</TargetFrameworks>
-    <VersionSuffix>alpha</VersionSuffix>
     <IsAotCompatible Condition="$([MSBuild]::IsTargetFrameworkCompatible('$(TargetFramework)', 'net7.0'))">true</IsAotCompatible>
+    <VersionSuffix>alpha</VersionSuffix>
   </PropertyGroup>
 
   <!-- IMPORT NUGET PACKAGE SHARED PROPERTIES -->

diff --git a/dotnet/src/Connectors/Connectors.Onnx/OnnxRuntimeGenAIChatCompletionService.cs b/dotnet/src/Connectors/Connectors.Onnx/OnnxRuntimeGenAIChatCompletionService.cs
@@ -25,7 +25,6 @@ public sealed class OnnxRuntimeGenAIChatCompletionService : IChatCompletionServi
     private readonly JsonSerializerOptions? _jsonSerializerOptions;
     private Model? _model;
     private Tokenizer? _tokenizer;
-
     private Dictionary<string, object?> AttributesInternal { get; } = new();
 
     /// <summary>

diff --git a/dotnet/src/Connectors/Connectors.Onnx/OnnxRuntimeGenAIPromptExecutionSettings.cs b/dotnet/src/Connectors/Connectors.Onnx/OnnxRuntimeGenAIPromptExecutionSettings.cs
@@ -11,6 +11,7 @@ namespace Microsoft.SemanticKernel.Connectors.Onnx;
 /// <summary>
 /// OnnxRuntimeGenAI Execution Settings.
 /// </summary>
+[JsonNumberHandling(JsonNumberHandling.AllowReadingFromString)]
 public sealed class OnnxRuntimeGenAIPromptExecutionSettings : PromptExecutionSettings
 {
     /// <summary>