Jump to content

Connect Leeroopedia MCP: Equip your AI agents to search best practices, build plans, verify code, diagnose failures, and look up hyperparameter defaults.

Environment:Dotnet Machinelearning Platform Architecture Support

From Leeroopedia


Knowledge Sources
Domains Infrastructure, Platform_Support
Last Updated 2026-02-09 11:00 GMT

Overview

Platform and CPU architecture requirements for ML.NET, with x64 as primary, partial ARM64/Apple M1 support, and limited 32-bit Windows support.

Description

ML.NET operates across multiple platforms and architectures with varying levels of component support. The x64 architecture has full support across all platforms. ARM64/Apple M1 support is available for training and inference but excludes several native-dependent components. 32-bit Windows has limited support excluding TensorFlow and LightGBM. Blazor WebAssembly has restricted support excluding most native components.

Usage

Use this environment definition to determine which ML.NET components are available on your target deployment platform. Check the compatibility matrix before selecting trainers or transforms in your pipeline to avoid runtime failures on non-x64 architectures.

System Requirements

Category Requirement Notes
OS Windows 10+, Linux (kernel 5.15+ recommended), macOS 10.13+ Cross-platform
Architecture x64 (primary), ARM64, x86 (Windows only) Component availability varies
RAM 2GB minimum More for large datasets

Dependencies

Linux Specific

  • `libc` >= 2.23 (required for ONNX Runtime and TorchSharp)

macOS Specific

  • macOS deployment target: 13.0
  • Homebrew for package management recommended

Credentials

No credentials required for platform support.

Quick Install

# Check your architecture
dotnet --info | grep "RID"

# Verify libc version on Linux (must be >= 2.23 for ONNX/TorchSharp)
ldd --version

Code Evidence

Architecture detection for native code from `src/Microsoft.ML.FastTree/Dataset/IntArray.cs:40`:

public static bool UseFastTreeNative => RuntimeInformation.ProcessArchitecture == Architecture.X64 ||
                                         RuntimeInformation.ProcessArchitecture == Architecture.X86;

OneDal x64-only check from `src/Microsoft.ML.OneDal/OneDalUtils.cs:24`:

if (System.Runtime.InteropServices.RuntimeInformation.ProcessArchitecture ==
    System.Runtime.InteropServices.Architecture.X64)

ONNX Runtime libc check from `test/Microsoft.ML.TestFrameworkCommon/Attributes/OnnxFactAttribute.cs:28-29`:

// ONNX Runtime requires Linux with libc >= 2.23

TorchSharp 64-bit requirement from `test/Microsoft.ML.TestFrameworkCommon/Attributes/TorchSharpFactAttribute.cs:16`:

// TorchSharp: 64-bit only, requires libc >= 2.23 on Linux

ARM module exclusion from `Directory.Build.targets:33-38`:

<!-- Excluded on ARM: MklImports, CpuMathNative, FastTreeNative, SymSgdNative, MklProxyNative -->

Common Errors

Error Message Cause Solution
`DllNotFoundException: FastTreeNative` Running on ARM64 where FastTreeNative is not available Use managed-only trainers (SDCA, LightGBM if compiled for ARM)
`GLIBC_2.23 not found` Linux libc version too old Upgrade to Ubuntu 16.04+ or equivalent with libc >= 2.23
`System.PlatformNotSupportedException` Component not supported on current platform Check platform limitations matrix and switch to supported component

Compatibility Notes

Component x64 ARM64/Apple M1 32-bit Windows Blazor WASM
Symbolic SGD Supported Not supported Supported Not supported
TensorFlow Supported Not supported Not supported Not supported
OLS Regression Supported Not supported Supported Not supported
TimeSeries SSA Supported Not supported Supported Not supported
TimeSeries SrCNN Supported Not supported Supported Not supported
ONNX Supported Inference only Supported Not supported
LightGBM Supported Can compile from source Not supported Not supported
LDA Supported Supported Supported Not supported
Matrix Factorization Supported Supported Supported Not supported

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment