Class TensorRtOnnxModelSupportReport
- Namespace
- JYPPX.TensorRtSharp
- Assembly
- JYPPX.TensorRtSharp.dll
Describes whether TensorRT reports support for an ONNX model and its parser subgraphs. 描述 TensorRT 是否支持指定 ONNX 模型以及 parser 子图支持情况。
public sealed class TensorRtOnnxModelSupportReport
- Inheritance
-
TensorRtOnnxModelSupportReport
- Inherited Members
Constructors
TensorRtOnnxModelSupportReport(bool, long, long, IReadOnlyList<TensorRtOnnxSubgraphSupportInfo>)
Creates an ONNX model support report. 创建 ONNX 模型支持性报告。
public TensorRtOnnxModelSupportReport(bool isSupported, long supportedSubgraphCount, long unsupportedSubgraphCount, IReadOnlyList<TensorRtOnnxSubgraphSupportInfo> subgraphs)
Parameters
isSupportedboolWhether TensorRT reports the full model as supported. TensorRT 是否报告完整模型受支持。
supportedSubgraphCountlongNumber of supported subgraphs. 受支持子图数量。
unsupportedSubgraphCountlongNumber of unsupported subgraphs. 不受支持子图数量。
subgraphsIReadOnlyList<TensorRtOnnxSubgraphSupportInfo>Per-subgraph support details. 每个子图的支持性详情。
Properties
IsSupported
Gets whether TensorRT reports the whole ONNX model as supported. 获取 TensorRT 是否报告整个 ONNX 模型受支持。
public bool IsSupported { get; }
Property Value
Subgraphs
Gets the per-subgraph support details. 获取每个子图的支持性详情。
public IReadOnlyList<TensorRtOnnxSubgraphSupportInfo> Subgraphs { get; }
Property Value
SupportedSubgraphCount
Gets the number of supported subgraphs reported by TensorRT. 获取 TensorRT 报告的受支持子图数量。
public long SupportedSubgraphCount { get; }
Property Value
UnsupportedSubgraphCount
Gets the number of unsupported subgraphs reported by TensorRT. 获取 TensorRT 报告的不受支持子图数量。
public long UnsupportedSubgraphCount { get; }
Property Value
Methods
ToString()
Returns a compact support summary. 返回紧凑的支持性摘要。
public override string ToString()
Returns
ToSummary()
Converts this report into a compact pointer-free summary. 将当前报告转换为紧凑、无指针逃逸的摘要。
public TensorRtOnnxModelSupportSummary ToSummary()
Returns
- TensorRtOnnxModelSupportSummary
A compact ONNX model support summary. 紧凑 ONNX 模型支持性摘要。
Remarks
This method only reads copied managed values that are already present in the report. It does not call TensorRT, does not expose parser-owned tensor pointers, and does not promote the report to runtime execution or package-consumer proof. 此方法只读取报告中已经复制到托管侧的值;不会调用 TensorRT、不会暴露 parser 拥有的 tensor 指针, 也不会将报告晋级为 runtime execution 或 package-consumer proof。