GpuDiagnosticLevel

Enums · AiDotNet.Configuration

Verbosity level for GPU backend diagnostic output.

For Beginners

Think of this like the log level in any logging framework: Silent is OFF, Minimal is “just the important stuff”, Verbose is “tell me everything”.

How It Works

Addresses github.com/ooples/AiDotNet#1122. AiDotNet’s GPU backends (OpenCL, HIP, CUDA) emit status messages during device discovery, kernel compilation, and availability checks. Historically these were always written to WriteLine, producing ~40 lines of output on every AiModelBuilder.BuildAsync(). This enum gives applications fine-grained control over that output.

Fields

Field Summary
GpuDiagnosticLevel Minimal Only critical GPU backend output — device selected, compilation failures, OpenCL DLL-not-found errors.
GpuDiagnosticLevel Silent No GPU backend output written to Console or any sink.
GpuDiagnosticLevel Verbose All GPU backend diagnostic output — device discovery, every kernel compilation step, OpenCL platform queries, GEMM tuning progress.
int value__