UAIX.LmRuntime / Package guide

UAIX.LmRuntime.Backends.Cuda.LegacyK80.win-x64

RID-specific NuGet package slot for Tesla K80 legacy CUDA native assets under runtimes/win-x64/native/.

Required For the separate Windows x64 Tesla K80 compute-capability 3.7 native-asset package slot

UAIX.LmRuntime.Backends.Cuda.LegacyK80.win-x64

RID-specific NuGet package slot for Tesla K80 legacy CUDA native assets under runtimes/win-x64/native/.

Overview

This package reserves the standard NuGet runtime asset path for the separate Windows x64 Tesla K80 compute-capability 3.7 native-asset package slot. It has no managed API surface. The supplied package source contains an asset manifest and no CUDA inference binary, so a package reference must not be treated as proof of native execution. Deployment and probe evidence remain required.

Who should use it Deployment engineers and native adapter authors separating Tesla K80 legacy CUDA assets from managed backend registration.

Install

.NET CLI

dotnet add package UAIX.LmRuntime.Backends.Cuda.LegacyK80.win-x64

Project file

<PackageReference Include="UAIX.LmRuntime.Backends.Cuda.LegacyK80.win-x64" />

Version policy: The documentation deliberately omits UAIX.LmRuntime package version numbers. Resolve and pin versions through your normal dependency-management and lock-file process.

Review the current package metadata, frameworks, dependencies, and downloads on NuGet ↗

Direct package dependencies

None within the UAIX.LmRuntime package family.

Package role and boundaries

Required For the separate Windows x64 Tesla K80 compute-capability 3.7 native-asset package slot

You are packaging or validating the separate Windows x64 Tesla K80 compute-capability 3.7 native-asset package slot under the win-x64 NuGet RID path.
You need the native-asset package identity to remain separate from managed backend registration.
You are testing output layout and future adapter probing without making an execution claim.

Boundary

It contains no managed runtime API.
The supplied source round embeds no CUDA inference binary.
It does not register Backends.Cuda and does not make the CUDA probe available.

RID-specific placement

Packable files under native/* are assigned to runtimes/win-x64/native/ by the project.

Separate package identity

Managed registration, modern CUDA assets, and Tesla K80 legacy assets are distinct package and evidence boundaries.

Execution requires proof

A usable adapter must prove file presence, loadability, compatible runtime libraries, device support, and execution behavior.

Key types

These are the main entry points for this package. The generated reference below includes every documented type and member represented by public package XML documentation.

No managed typesThis package is a native-asset identity or deployment slot. It is consumed through a managed backend package rather than called from C# directly.

Coding examples

Examples use public package signatures documented on LMRuntime.com. Model paths, hashes, byte counts, prompts, and host-specific identifiers remain application inputs.

Reference the package only for win-x64 publishing

Keep the RID-specific package reference scoped to the Windows x64 deployment target.

App.csproj

<Project Sdk="Microsoft.NET.Sdk">
  <PropertyGroup>
    <TargetFramework>net9.0</TargetFramework>
    <RuntimeIdentifier>win-x64</RuntimeIdentifier>
  </PropertyGroup>
  <ItemGroup Condition="'$(RuntimeIdentifier)' == 'win-x64'">
    <PackageReference Include="UAIX.LmRuntime.Backends.Cuda.LegacyK80.win-x64" />
  </ItemGroup>
</Project>

Boundary: A conditional package reference controls restore topology; it does not verify the content or loadability of a native asset.

Publish and inspect the native output boundary

Create the RID-specific output and enumerate native files before a deployment claim is made.

terminal

dotnet publish --configuration Release --runtime win-x64

# Inspect the published output with your normal release tooling.
# A package ID or directory alone is not proof of a usable CUDA engine.

Boundary: The supplied package source contains only an asset manifest. Validate the concrete package version and output contents in the release pipeline.

Keep CUDA registration and native assets as separate checks

Probe the managed CUDA backend after deployment rather than assuming that restore activated GPU execution.

CudaDeploymentProbe.cs

using UAIX.LmRuntime.Acceleration;
using UAIX.LmRuntime.Backends.Cuda;

var backend = new CudaRuntimeBackend();
RuntimeBackendProbeResult probe = await backend.ProbeAsync(
    new RuntimeBackendOptions
    {
        RequestedRuntimeIdentifier = "win-x64",
        RequireNativeAssets = true,
        NativeAssetDirectory = AppContext.BaseDirectory
    });

if (!probe.IsAvailable)
{
    foreach (string diagnostic in probe.Diagnostics)
    {
        Console.Error.WriteLine(diagnostic);
    }
}

Boundary: The supplied diagnostic CUDA backend does not load native files. A production native adapter must provide the proof before IsAvailable can be true.

Open this package on NuGet Compare the full package family →

Generated API reference

Expand a type to review its documented fields, properties, constructors, methods, parameter descriptions, and return descriptions. Browser Find also works across the closed detail elements.

No managed API surfaceThis package carries a native-asset identity or deployment slot and contains no documented managed types. Review its package guide and current NuGet metadata for the exact asset boundary.

Frequently asked questions

Why is there no API reference?

This is a native-asset package slot with no managed assembly API.

Does the current package prove CUDA execution?

No. The supplied source asset manifest explicitly states that no CUDA inference binary is embedded.

Why is the K80 package separate?

Tesla K80 compute capability 3.7 can require a legacy CUDA and sm_37-specific build that must not be mixed into the modern asset set.