What's new in .NET 8 RC 2 [WIP] #8440
Comments
.NET Libraries READMEsOne of the top customer problems that package consumers are facing is lack of documentation. As such, we are driving an effort to increase the adoption and quality of NuGet package READMEs. The README file is an essential part of your package as it provides important information to users and helps them understand what the package is and what it does quickly. Also, README is the first things for users when they view your package on NuGet.org and soon other tooling, it is crucial for package authors to write and include high-quality READMEs for their packages. Now there are READMEs for the following Microsoft packages: Priority 1 (.NET 8 RC 2)The below list is sorted by download count.
Priority 2 (.NET 8 / .NET 9)This list is sorted alphabetically.
Note: For the blog writer, please only include the packages that have ✅ as this work will continue to .NET 9. |
|
@JonDouglas: I have 404 in all URLs in your message: |
See the parent issue for more information. Likely the |
MSBuild: Simple CLI-based project evaluationMSBuild is a very powerful platform, and is great at integrating data and tools from other ecosystems into its view of the world. It's historically not been so good at making the data that it has available to the broader worlds of scripting and tools. In the past, tools authors would have to do things like injecting their own MSBuild code to read certain Properties, Items, or Target Outputs into files, and then parse those files. This is work that is very error prone. In .NET 8, the MSBuild team has shipped a new feature that makes it easier to incorporate data from MSBuild into your scripts or tools. Let's take a look at a sample to see what I mean. >dotnet publish --getProperty:OutputPath
bin\Release\net8.0\In this simple example, we've used the > dotnet publish -p PublishProfile=DefaultContainer --getProperty:GeneratedContainerDigest --getProperty:GeneratedContainerConfiguration
{
"Properties": {
"GeneratedContainerDigest": "sha256:ef880a503bbabcb84bbb6a1aa9b41b36dc1ba08352e7cd91c0993646675174c4",
"GeneratedContainerConfiguration": "{\u0022config\u0022:{\u0022ExposedPorts\u0022:{\u00228080/tcp\u0022:{}},\u0022Labels\u0022:{\u0022org.opencontainers.image.created\u0022:\u00222023-10-02T18:20:01.6356453Z\u0022,\u0022org.opencontainers.artifact.created\u0022:\u00222023-10-02T18:20:01.6356453Z\u0022,\u0022org.opencontainers.artifact.description\u0022:\u0022A project that demonstrates publishing to various container registries using just\\r\\n the .NET SDK\u0022,\u0022org.opencontainers.image.description\u0022:\u0022A project that demonstrates publishing to various container registries using just\\r\\n the .NET SDK\u0022,\u0022org.opencontainers.image.authors\u0022:\u0022Chet Husk\u0022,\u0022org.opencontainers.image.url\u0022:\u0022https://github.com/baronfel/sdk-container-demo\u0022,\u0022org.opencontainers.image.documentation\u0022:\u0022https://github.com/baronfel/sdk-container-demo\u0022,\u0022org.opencontainers.image.version\u0022:\u00221.0.0\u0022,\u0022org.opencontainers.image.licenses\u0022:\u0022MIT\u0022,\u0022org.opencontainers.image.title\u0022:\u0022.NET SDK 7 Container Demo\u0022,\u0022org.opencontainers.image.base.name\u0022:\u0022mcr.microsoft.com/dotnet/aspnet:8.0.0-rc.1\u0022},\u0022Env\u0022:[\u0022PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin\u0022,\u0022APP_UID=1654\u0022,\u0022ASPNETCORE_HTTP_PORTS=8080\u0022,\u0022DOTNET_RUNNING_IN_CONTAINER=true\u0022,\u0022DOTNET_VERSION=8.0.0-rc.1.23419.4\u0022,\u0022ASPNET_VERSION=8.0.0-rc.1.23421.29\u0022],\u0022WorkingDir\u0022:\u0022/app\u0022,\u0022Entrypoint\u0022:[\u0022dotnet\u0022,\u0022sdk-container-demo.dll\u0022],\u0022User\u0022:\u00221654\u0022},\u0022created\u0022:\u00222023-10-02T18:20:04.6914310Z\u0022,\u0022rootfs\u0022:{\u0022type\u0022:\u0022layers\u0022,\u0022diff_ids\u0022:[\u0022sha256:d310e774110ab038b30c6a5f7b7f7dd527dbe527854496bd30194b9ee6ea496e\u0022,\u0022sha256:379caff0dd639afb033e114cb8da17c334a36d0c6c01bb4cf5f2d1a811968742\u0022,\u0022sha256:80627b9413613b9703eec6adc7a3a751ac1b7571041c77899456345f823ef63a\u0022,\u0022sha256:6231bc2ccd45860760c09a50d6d059aa4b6aa357e41e6d06f4394f24176f203d\u0022,\u0022sha256:56fa0b94fd5e406124a3d070ec79998698ddea2e635bf53bbf106dc86aeaa240\u0022,\u0022sha256:272eedde5582036a7f26fe5d069f4ba328ba7a5c6be30f6dcbee9838224df148\u0022,\u0022sha256:4b8ab71658cccccfaf8979b1025f3ed1b12e936a448dcd13b9ab4f7709f31357\u0022]},\u0022architecture\u0022:\u0022amd64\u0022,\u0022os\u0022:\u0022linux\u0022,\u0022history\u0022:[{\u0022created\u0022:\u00222023-09-20T04:55:40.8154909Z\u0022,\u0022created_by\u0022:\u0022/bin/sh -c #(nop) ADD file:a1398394375faab8dd9e1e8d584eea96c750fb57ae4ffd2b14624f1cf263561b in / \u0022},{\u0022created\u0022:\u00222023-09-20T04:55:41.1203677Z\u0022,\u0022created_by\u0022:\u0022/bin/sh -c #(nop) CMD [\\u0022bash\\u0022]\u0022,\u0022empty_layer\u0022:true},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:34.5265068Z\u0022,\u0022created_by\u0022:\u0022ENV APP_UID=1654 ASPNETCORE_HTTP_PORTS=8080 DOTNET_RUNNING_IN_CONTAINER=true\u0022,\u0022empty_layer\u0022:true},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:34.5265068Z\u0022,\u0022created_by\u0022:\u0022RUN /bin/sh -c apt-get update \\u0026\\u0026 apt-get install -y --no-install-recommends ca-certificates libc6 libgcc-s1 libicu72 libssl3 libstdc\\u002B\\u002B6 tzdata zlib1g \\u0026\\u0026 rm -rf /var/lib/apt/lists/* # buildkit\u0022},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:34.8536814Z\u0022,\u0022created_by\u0022:\u0022RUN /bin/sh -c groupadd --gid=$APP_UID app \\u0026\\u0026 useradd -l --uid=$APP_UID --gid=$APP_UID --create-home app # buildkit\u0022},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:36.8764982Z\u0022,\u0022created_by\u0022:\u0022ENV DOTNET_VERSION=8.0.0-rc.1.23419.4\u0022,\u0022empty_layer\u0022:true},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:36.8764982Z\u0022,\u0022created_by\u0022:\u0022COPY /dotnet /usr/share/dotnet # buildkit\u0022},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:37.1287440Z\u0022,\u0022created_by\u0022:\u0022RUN /bin/sh -c ln -s /usr/share/dotnet/dotnet /usr/bin/dotnet # buildkit\u0022},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:39.8264990Z\u0022,\u0022created_by\u0022:\u0022ENV ASPNET_VERSION=8.0.0-rc.1.23421.29\u0022,\u0022empty_layer\u0022:true},{\u0022comment\u0022:\u0022buildkit.dockerfile.v0\u0022,\u0022created\u0022:\u00222023-09-20T12:13:39.8264990Z\u0022,\u0022created_by\u0022:\u0022COPY /shared/Microsoft.AspNetCore.App /usr/share/dotnet/shared/Microsoft.AspNetCore.App # buildkit\u0022},{\u0022author\u0022:\u0022.NET SDK\u0022,\u0022created\u0022:\u00222023-10-02T18:20:04.6914069Z\u0022,\u0022created_by\u0022:\u0022.NET SDK Container Tooling, version 8.0.100-dev\u0022}]}"
}
}Here we've requested two properties, and received a JSON object with a The same behavior holds for MSBuild Items, which can be requested by Item Type using the >dotnet publish -p PublishProfile=DefaultContainer --getItem:ContainerImageTags
{
"Items": {
"ContainerImageTags": [
{
"Identity": "latest",
.... other MSBuild Item Metadata elided ....
]
}
}Finally, we also support retrieving the outputs of a given Target that's been run using dotnet publish -p PublishProfile=DefaultContainer --getTargetResult:ComputeContainerBaseImage
{
"TargetResults": {
"ComputeContainerBaseImage": {
"Result": "Success",
"Items": [
{
"Identity": "mcr.microsoft.com/dotnet/aspnet:8.0.0-rc.1",
.... other MSBuild Item Metadata elided ....
}
]
}
}
}We think this will really make it easier for MSBuild to integrate into other tooling, so we're excited to see what you all do with it! |
SDK Container Publish: Publish to tar.gz archiveThe SDK has been able to publish containers to local container tools like Docker and Podman, as well as Remote Container Registries like Azure Container Registry, Amazon's Elastic Container Registry, and Docker Hub for quite some time now. However, not all workflows are so straightforward. Some teams may prefer to run scanning tools over their images before pushing them, for example. To help support these workflows, community member @Danielku15 implemented an awesome new feature for the SDK Container Publish tooling - the ability to create a container directly as a tar.gz archive. Once this archive is made, it can be moved, scanned, or loaded into a local Docker toolchain - whatever your needs may be! Let's take a look at how simple it is: >dotnet publish -p PublishProfile=DefaultContainer -p ContainerArchiveOutputPath=./images/sdk-container-demo.tar.gz
MSBuild version 17.8.0+6cdef4241 for .NET
Determining projects to restore...
All projects are up-to-date for restore.
C:\Program Files\dotnet\sdk\8.0.100-rc.2.23477.19\Sdks\Microsoft.NET.Sdk\targets\Microsoft.NET.RuntimeIdentifierInference.targets(311,5): message NETSDK1057: You are using a preview version of .NET. See: http
s://aka.ms/dotnet-support-policy [D:\code\sdk-container-demo\src\sdk-container-demo\sdk-container-demo.csproj]
sdk-container-demo -> D:\code\sdk-container-demo\src\sdk-container-demo\bin\Release\net8.0\sdk-container-demo.dll
sdk-container-demo -> D:\code\sdk-container-demo\src\sdk-container-demo\bin\Release\net8.0\publish\
Building image 'sdk-container-demo' with tags 'latest' on top of base image 'mcr.microsoft.com/dotnet/aspnet:8.0.0-rc.1'.
Pushed image 'sdk-container-demo:latest' to local archive at 'D:\code\sdk-container-demo\src\sdk-container-demo\images\sdk-container-demo.tar.gz'.Here I've added one new property to my publish command: Once you have a tar.gz, you can move it wherever you need, or simply load it into Docker using >docker load --input .\images\sdk-container-demo.tar.gz
5d6d1d62da4b: Loading layer [==================================================>] 9.442MB/9.442MB
Loaded image: sdk-container-demo:latestJust like that my image is available for me to run via Docker. We hope this new capability enables new workflows for more teams - let us know what you think at our repo! |
|
@JonDouglas and team - this next one is an advisory notice that SDK/Roslyn teams think is big enough to call-out somehow on the post. Happy to chat about it more. Important .NET SDK 8.0.100-rc.2 must be used with Visual Studio 17.8 preview 3 due to a dependency error with Razor tooling. See this SDK Announcement for more details. |
Introducing Tensor Primitives for .NETOver the past few years, .NET has made significant refinements to numeric support in an effort to optimize data-intensive workloads like those of AI and Machine Learning. Starting with the addition of hardware intrinsics in .NET Core 3, we provided .NET developers with access to hardware specific instructions. This enabled .NET applications to utilize modern hardware capabilities like vector instructions which are critical to AI workloads. The next step in this effort was Generic Math, which was introduced in .NET 6 as a preview feature and marked stable in .NET 7. Generic Math provides more type-safe simplified numeric operations. This feature eliminates the need for creating numerous nearly identical implementations to cater to different numeric types, thus simplifying the code and making it more maintainable. Tensor Primitives is the next step in the evolution of Numerics for AI in .NET by building on top of hardware intrinsics and Generic Math, and this set of features is included in .NET 8 starting with RC2. Tensor Primitives is short for For more details, see the Future of Numerics and AI proposal. Semantic search exampleAI workloads, such as semantic search or Retrieval Augmented Generation (RAG), enhance the natural language capabilities of Large Language Models like ChatGPT by incorporating relevant data into prompts. In these tasks, operations on vectors, such as cosine similarity, are essential for identifying the most relevant data to answer a question. Let's imagine you have a movie database which contains movie titles and embeddings. Embeddings are beyond the scope of this post, but they're ways of encoding semantic information as an array of numbers. In this case, they represent a brief synopsis of these movie plots and they've been precomputed. However, if you'd like you can use models like those from Azure OpenAI to generate embeddings. var movies = new [] {
new {Title="The Lion King", Embedding= new [] {0.10022575f, -0.23998135f}},
new {Title="Inception", Embedding= new [] {0.10327095f, 0.2563685f}},
new {Title="Toy Story", Embedding= new [] {0.095857024f, -0.201278f}},
new {Title="Pulp Function", Embedding= new [] {0.106827796f, 0.21676421f}},
new {Title="Shrek", Embedding= new [] {0.09568083f, -0.21177962f}}
};Let's say that you wanted to search for family friendly movies using the search term "A movie that's fun for the whole family". The embedding for that query might look as follows. var queryEmbedding = new[] {0.12217915f, -0.034832448f };At this point, to see which movies in the database more closely match my query, I can compute the distance using a distance function like cosine similarity. Before Tensor PrimitivesBefore
Using external dependenciesThe easiest way for you to use any of these operations was to use existing libraries such as TorchSharp, TensorFlow.NET, Math.NET Numerics, ML.NET, or Semantic Kernel. If the application you were building heavily relied on the functionality provided by those libraries, it made sense to add them to your project. However, if all you needed was a handful of methods from those libraries, taking a dependency on those libraries introduced additional overhead. using TorchSharp;
using static TorchSharp.torch.nn.functional;
var top3MoviesTorchSharp =
movies
.Select(movie =>
new {
Title=movie.Title,
Embedding=movie.Embedding,
Similarity=cosine_similarity(torch.tensor(queryEmbedding), torch.tensor(movie.Embedding), 0L).item<float>()})
.OrderByDescending(movies => movies.Similarity)
.Take(3);This is what the code might look like for measuring the similarity of my query to my movie collection. After sorting by most similar (higher is more similar) and taking the top 3 movies most likely to be "family-friendly", the output is the following:
As you can see from this example, movies such as Pulp Fiction and Inception are not part of the results because they're not considered "family-friendly" movies. While TorchSharp makes it easy for you to calculate cosine similarity, adding it as a dependency for this one function may not make sense. Writing your own implementationIf you didn't want to take on another external dependency just to use a few operations in your application, the other alternative was to implement your own. This is what many of the libraries mentioned previously have done. While this is a viable path that offers you the most control over your code, this meant you were writing framework-level code rather than focusing on the competitive advantages of your application or business. Most likely your code might've been a naive implementation of the operation which didn't take full advantage of the runtime's capabilities for hardware optimizations. public float CosineSimilarityCustom(ReadOnlySpan<float> vector1, ReadOnlySpan<float> vector2)
{
if (vector1.Length != vector2.Length)
throw new ArgumentException("Vectors must have the same length");
float dotProduct = 0f;
float magnitude1 = 0f;
float magnitude2 = 0f;
for (int i = 0; i < vector1.Length; i++)
{
dotProduct += vector1[i] * vector2[i];
magnitude1 += vector1[i] * vector1[i];
magnitude2 += vector2[i] * vector2[i];
}
magnitude1 = MathF.Sqrt(magnitude1);
magnitude2 = MathF.Sqrt(magnitude2);
if (magnitude1 == 0 || magnitude2 == 0)
return 0; // handle the case where one or both vectors have zero magnitude
return dotProduct / (magnitude1 * magnitude2);
}This is what a custom implementation of cosine similarity might look like. You'd then use this as follows. var top3MoviesCustom =
movies
.Select(movie =>
new {
Title=movie.Title,
Embedding=movie.Embedding,
Similarity=CosineSimilarityCustom(queryEmbedding, movie.Embedding)})
.OrderByDescending(movies => movies.Similarity)
.Take(3);The results of this code are the same as those from the TorchSharp example. While there are no external dependencies in this case, cosine similarity now becomes a set of framework-level code you need to maintain. After Tensor Primitives
using System.Numerics.Tensors;
var top3MoviesTensorPrimitives =
movies
.Select(movie =>
new {
Title=movie.Title,
Embedding=movie.Embedding,
Similarity=TensorPrimitives.CosineSimilarity(queryEmbedding, movie.Embedding)})
.OrderByDescending(movies => movies.Similarity)
.Take(3);Similarly, for libraries who make use of these operations such as Semantic Kernel and ML.NET, they can replace many of their existing implementation with TensorPrimitives so they can focus on their competitive advantages and deliver features faster. Current StateCurrently, Tensor Primitives provides vectorized implementations for operations like:
For a complete list of operations, see Tensor Primitives work items tracking issue. Suggested UsageIn initial tests, we've observed that the performance efficiencies this package introduces are more noticeable on tensors with a large number of elements. What's next?Over the next few months, we plan to:
Get started with Tensor Primitives todayIn your .NET project, add a reference to the latest Give us feedback and file any issues in the dotnet/runtime repo. |
|
@luisquintanilla shouldn't it be |
What's new in .NET 8 RC 2
This issue is for teams to highlight work for the community that will release in .NET 8 RC 2
To add content, use a new conversation entry. The entry should include the team name and feature title as the first line shown in the template below.
Required
Optional
Below are three additional items to consider. These will help the .NET 8 blog team and the community throughout the release.
Index of .NET 8 releases
Preview 1: #8133
Preview 2: #8134
Preview 3: #8135
Preview 4: #8234
Preview 5: #8436
Preview 6: #8437
Preview 7: #8438
RC 1: #8439
RC 2: #8440
The text was updated successfully, but these errors were encountered: