OnnxRuntimeSessionOptions
Inheritance: java.lang.Object
public class OnnxRuntimeSessionOptions
Configuration options for creating ONNX InferenceSession. We recommend keeping the optimized defaults unless you are absolutely certain about the modifications. For technical details, refer to ONNX Runtime documentation.
Constructors
| Constructor | Description |
|---|---|
| OnnxRuntimeSessionOptions() |
Fields
| Field | Description |
|---|---|
| enableCpuMemArena | Enables or disables the CPU memory arena allocator used by ONNX Runtime. |
| enableMemoryPattern | Enables or disables memory pattern optimization for input tensors. |
| executionMode | Execution mode for the session. |
| graphOptimizationLevel | Graph optimization level for the session. |
| interOpNumThreads | Number of threads for running multiple operations in parallel. |
| intraOpNumThreads | Number of threads for a single operation. |
OnnxRuntimeSessionOptions()
public OnnxRuntimeSessionOptions()
enableCpuMemArena
public static boolean enableCpuMemArena
Enables or disables the CPU memory arena allocator used by ONNX Runtime. When enabled, memory is pooled and reused for better performance, but may lead to increased memory consumption in multi-threaded scenarios. Disable to reduce peak memory usage at the cost of performance.
enableMemoryPattern
public static boolean enableMemoryPattern
Enables or disables memory pattern optimization for input tensors. When enabled, ONNX Runtime caches memory allocation patterns for faster execution, but may increase memory usage for dynamic input shapes. Disable if inputs vary significantly or to reduce memory footprint.
executionMode
public static ExecutionModeOnnx executionMode
Execution mode for the session. By default, operators are executed concurrently, whenever possible.
graphOptimizationLevel
public static GraphOptimizationLevelOnnx graphOptimizationLevel
Graph optimization level for the session. By default, all available optimizations are enabled for maximum performance.
interOpNumThreads
public static int interOpNumThreads
Number of threads for running multiple operations in parallel. If sequential execution is enabled, this value is ignored.
intraOpNumThreads
public static int intraOpNumThreads
Number of threads for a single operation.