Class TensorRtEngineDeploymentSnapshot
- Namespace
- JYPPX.TensorRtSharp
- Assembly
- JYPPX.TensorRtSharp.dll
Captures TensorRT engine deployment metadata that is useful before binding buffers and enqueueing inference. 捕获在绑定缓冲区与提交推理前有价值的 TensorRT engine 部署元数据。
public sealed class TensorRtEngineDeploymentSnapshot
- Inheritance
-
TensorRtEngineDeploymentSnapshot
- Inherited Members
Properties
AuxiliaryStreamCount
Gets the auxiliary stream count requested by the engine. 获取 engine 请求的辅助 CUDA stream 数量。
public int AuxiliaryStreamCount { get; }
Property Value
Capability
Gets the engine capability recorded by TensorRT. 获取 TensorRT 记录的 engine capability。
public TensorRtEngineCapability Capability { get; }
Property Value
DeviceMemorySizeInBytes
Gets TensorRT's base device-memory requirement. 获取 TensorRT 报告的基础 device memory 需求。
public ulong DeviceMemorySizeInBytes { get; }
Property Value
DeviceMemorySizeV2InBytes
Gets TensorRT 11 V2 device-memory requirement. 获取 TensorRT 11 V2 device memory 需求。
public ulong DeviceMemorySizeV2InBytes { get; }
Property Value
Diagnostics
Gets non-fatal diagnostics collected while building the snapshot. 获取构建快照时收集的非致命诊断信息。
public IReadOnlyList<string> Diagnostics { get; }
Property Value
EngineName
Gets the TensorRT engine name. 获取 TensorRT engine 名称。
public string EngineName { get; }
Property Value
HardwareCompatibilityLevel
Gets TensorRT 11 hardware compatibility level recorded by the engine. 获取 TensorRT 11 engine 记录的硬件兼容级别。
public TensorRtHardwareCompatibilityLevel HardwareCompatibilityLevel { get; }
Property Value
IOTensorCount
Gets the number of named input/output tensors in the engine. 获取 engine 中命名输入/输出 tensor 的数量。
public int IOTensorCount { get; }
Property Value
IsRefittable
Gets whether TensorRT reports this engine as refittable. 获取 TensorRT 是否报告该 engine 可 refit。
public bool IsRefittable { get; }
Property Value
LayerCount
Gets the number of layers recorded in the engine. 获取 engine 中记录的 layer 数量。
public int LayerCount { get; }
Property Value
OptimizationProfileCount
Gets the number of optimization profiles available in the engine. 获取 engine 中可用的 optimization profile 数量。
public int OptimizationProfileCount { get; }
Property Value
ProfileDeviceMemorySizeInBytes
Gets profile-specific device-memory requirement. 获取指定 profile 的 device memory 需求。
public ulong ProfileDeviceMemorySizeInBytes { get; }
Property Value
ProfileDeviceMemorySizeV2InBytes
Gets profile-specific V2 device-memory requirement. 获取指定 profile 的 V2 device memory 需求。
public ulong ProfileDeviceMemorySizeV2InBytes { get; }
Property Value
ProfileIndex
Gets the optimization profile used when collecting profile-specific fields. 获取采集 profile 相关字段时使用的 optimization profile 索引。
public int ProfileIndex { get; }
Property Value
ProfileTensorValues
Gets copied profile tensor values snapshots collected for tensor/profile diagnostics. 获取为 tensor/profile 诊断采集的 copied profile tensor values 快照。
public IReadOnlyList<TensorRtEngineProfileTensorValuesSnapshot> ProfileTensorValues { get; }
Property Value
ProfilingVerbosity
Gets profiling verbosity recorded by the engine. 获取 engine 记录的 profiling verbosity。
public TensorRtProfilingVerbosity ProfilingVerbosity { get; }
Property Value
StreamableWeightsSizeInBytes
Gets streamable weights size reported by TensorRT 11. 获取 TensorRT 11 报告的可流式加载权重大小。
public long StreamableWeightsSizeInBytes { get; }
Property Value
StrippedWeightsSizeInBytes
Gets stripped weights size reported by TensorRT. 获取 TensorRT 报告的 stripped weights 大小。
public long StrippedWeightsSizeInBytes { get; }
Property Value
TacticSources
Gets tactic-source bit flags used by the engine. 获取 engine 使用的 tactic source 位标志。
public TensorRtTacticSources TacticSources { get; }
Property Value
Tensors
Gets per-tensor binding metadata collected from the engine. 获取从 engine 采集的逐 tensor 绑定元数据。
public IReadOnlyList<TensorRtEngineTensorBinding> Tensors { get; }
Property Value
TotalWeightsSizeInBytes
Gets total weights size reported by TensorRT. 获取 TensorRT 报告的总权重大小。
public long TotalWeightsSizeInBytes { get; }
Property Value
WeightStreamingAutomaticBudgetInBytes
Gets TensorRT's automatic weight-streaming budget. 获取 TensorRT 自动选择的权重流式加载预算。
public long WeightStreamingAutomaticBudgetInBytes { get; }
Property Value
WeightStreamingBudgetV2InBytes
Gets current weight-streaming budget. 获取当前权重流式加载预算。
public long WeightStreamingBudgetV2InBytes { get; }
Property Value
WeightStreamingScratchMemorySizeInBytes
Gets scratch-memory requirement for weight streaming. 获取权重流式加载所需的临时内存大小。
public long WeightStreamingScratchMemorySizeInBytes { get; }
Property Value
Methods
ToString()
Creates a compact one-line diagnostic summary. 创建紧凑的单行诊断摘要。
public override string ToString()
Returns
ToSummary()
Converts this copied engine deployment snapshot into a compact pointer-free summary. 将当前已复制 engine deployment 快照转换为紧凑、无指针逃逸的摘要。
public TensorRtEngineDeploymentSummary ToSummary()
Returns
Remarks
This method only reads managed snapshot values. It does not call TensorRT, expose native engine pointers, or promote local deployment diagnostics to runtime or package-consumer proof. 该方法只读取托管快照值;不会调用 TensorRT、暴露原生 engine 指针,也不会将本地部署诊断 晋级为 runtime 或 package-consumer proof。