Namespace JYPPX.TensorRtSharp
Classes
- BridgeProbeException
Thrown when the bridge reports a non-success status for a probe operation. 当 bridge 在探测操作中返回非成功状态时引发。
- TensorRtAttention
Represents a TensorRT 11 attention object owned by a network. 表示由 network 持有生命周期的 TensorRT 11 attention 对象。
- TensorRtBuilder
Managed wrapper around a TensorRT builder. TensorRT builder 的托管封装。
- TensorRtBuilderConfig
Managed wrapper around a TensorRT builder configuration. TensorRT builder 配置的托管封装。
- TensorRtBuilderConfigDeploymentSnapshot
Captures TensorRT builder configuration values that materially affect engine deployment. 捕获会实质影响 engine 部署结果的 TensorRT builder 配置值。
- TensorRtDependencyProbeReport
Non-throwing diagnostic report for native bridge, TensorRT, parser, CUDA, and cuDNN dependency resolution. 用于 native bridge、TensorRT、parser、CUDA 与 cuDNN 依赖解析的非抛异常诊断报告。
- TensorRtDims
Represents a TensorRT dimension vector. 表示一个 TensorRT 维度向量。
- TensorRtDims64
Represents TensorRT 11 dimensions whose extents are reported as 64-bit values. 表示 TensorRT 11 中以 64 位整数报告的维度信息。
- TensorRtEngine
Managed wrapper around a TensorRT engine. TensorRT engine 的托管封装。
- TensorRtEngineBindingReport
Summarizes TensorRT engine I/O metadata and optional execution-context readiness for deployment. 汇总 TensorRT engine I/O 元数据以及可选 execution-context 就绪状态，用于模型部署诊断。
- TensorRtEngineDeploymentSnapshot
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。
- TensorRtEngineInspector
Managed wrapper around a TensorRT engine inspector. TensorRT engine inspector 的托管封装。
- TensorRtEngineTensorBinding
Describes one TensorRT engine I/O tensor for deployment binding diagnostics. 描述一个 TensorRT engine I/O tensor 的部署绑定诊断信息。
- TensorRtEnvironmentProbe
Queries the current bridge state without requiring TensorRT inference to succeed. 在不要求 TensorRT 推理成功的前提下查询当前 bridge 状态。
- TensorRtEnvironmentSnapshot
High-level environment snapshot for the native TensorRT bridge. 原生 TensorRT bridge 的高层环境快照。
- TensorRtException
Base exception for TensorRT bridge operations. TensorRT bridge 操作的基础异常类型。
- TensorRtExecutionContext
Represents a managed TensorRT Tensor Rt Execution Context wrapper. 表示托管 TensorRT Tensor Rt Execution Context 包装器。
- TensorRtExecutionContextDeploymentSnapshot
Captures TensorRT execution-context state relevant to dynamic shapes, tensor addresses, and enqueue readiness. 捕获与动态 shape、tensor 地址绑定和 enqueue 就绪状态相关的 TensorRT execution-context 状态。
- TensorRtExecutionContextReadiness
Summarizes whether a TensorRT execution context has enough shape and address state for enqueue. 汇总 TensorRT execution context 是否已经具备 enqueue 所需的 shape 与地址绑定状态。
- TensorRtGlobalRuntimeVersion
Read-only TensorRT global runtime version details collected without creating a runtime. 无需创建 runtime 即可采集的 TensorRT 全局 runtime 版本信息。
- TensorRtHostMemory
Managed wrapper around TensorRT host memory. TensorRT host memory 的托管封装。
- TensorRtIfConditional
Represents a TensorRT if-conditional definition owned by a network. 表示由 network 持有生命周期的 TensorRT if-conditional 定义。
- TensorRtInferenceBindings
High-level TensorRT inference binding set for deployment-oriented enqueue flows. 面向模型部署 enqueue 流程的高层 TensorRT 推理绑定集。
- TensorRtInferenceBuffer
Describes one managed inference buffer bound to a TensorRT engine tensor. 描述一个绑定到 TensorRT engine tensor 的托管推理缓冲区。
- TensorRtInferenceExecutionSummary
Summarizes one TensorRT inference enqueue attempt. 汇总一次 TensorRT 推理 enqueue 操作。
- TensorRtLayer
Represents a managed TensorRT Tensor Rt Layer wrapper. 表示托管 TensorRT Tensor Rt Layer 包装器。
- TensorRtLayerTensorMetadata
Describes one TensorRT layer input or output tensor slot for deployment diagnostics. 描述 TensorRT layer 的一个输入或输出 tensor 槽位，用于部署诊断。
- TensorRtLogger
Managed wrapper around a TensorRT logger. TensorRT logger 的托管封装。
- TensorRtLoop
Represents a TensorRT loop definition owned by a network. 表示由 network 持有生命周期的 TensorRT loop 定义。
- TensorRtNativeDependencyInfo
Describes a native bridge, TensorRT, CUDA, cuDNN, or parser DLL discovered by a dependency probe. 描述依赖探针发现的 native bridge、TensorRT、CUDA、cuDNN 或 parser DLL。
- TensorRtNetworkDefinition
Represents a managed TensorRT Tensor Rt Network Definition wrapper. 表示托管 TensorRT Tensor Rt Network Definition 包装器。
- TensorRtOnnxModelSupportReport
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。
- TensorRtOnnxParser
Parses ONNX models into a TensorRT network definition. 将 ONNX 模型解析到 TensorRT network definition。
- TensorRtOnnxParserDiagnostic
Provides detailed diagnostics for one ONNX parser error reported by TensorRT. 表示 TensorRT ONNX 解析器报告的一条详细错误诊断信息。
- TensorRtOnnxSubgraphSupportInfo
Describes one ONNX parser subgraph support result. 描述一条 ONNX parser 子图支持性结果。
- TensorRtOptimizationProfile
Wraps a TensorRT optimization profile used to describe dynamic input ranges during engine building. 封装 TensorRT 优化配置文件，用于在构建引擎时描述动态输入的 min/opt/max 范围。
- TensorRtOptimizationProfileShapeRange
Represents the min/opt/max dimension range of a TensorRT optimization profile input. 表示 TensorRT 优化 profile 输入张量的 min/opt/max 维度范围。
- TensorRtOptimizationProfileShapeRange64
Describes a TensorRT 11 min/opt/max shape range using 64-bit dimension extents. 使用 64 位维度 extent 描述 TensorRT 11 的 min/opt/max shape 范围。
- TensorRtOptimizationProfileShapeValueRange
Represents the min/opt/max integer value range of a TensorRT shape tensor input. 表示 TensorRT shape tensor 输入的 min/opt/max 整数取值范围。
- TensorRtOptimizationProfileShapeValueRangeV2
Represents the min/opt/max 64-bit value range of a TensorRT shape tensor input. 表示 TensorRT shape tensor 输入的 min/opt/max 64 位整数取值范围。
- TensorRtParserErrorInfo
Provides fixed-buffer ONNX parser error information returned by TensorRT. 表示 TensorRT 返回的固定缓冲区 ONNX parser 错误信息。
- TensorRtPluginCreatorInfo
Read-only metadata for a TensorRT plugin creator. TensorRT plugin creator 的只读元数据。
- TensorRtPluginFieldInfo
Read-only metadata for a TensorRT plugin creator field. TensorRT plugin creator 字段的只读元数据。
- TensorRtPluginRegistryInventory
Read-only snapshot of a TensorRT plugin registry. TensorRT plugin registry 的只读快照。
- TensorRtRefitWeightsBuffer
Keeps TensorRT refit weights pinned while the native refitter may need to read them. 在原生 refitter 可能读取权重期间，保持 TensorRT refit 权重处于 pinned 状态。
- TensorRtRefitter
Provides a safe managed wrapper over TensorRT refitter operations. 提供 TensorRT refitter 操作的安全托管封装。
- TensorRtRuntime
Managed wrapper around a TensorRT runtime. TensorRT runtime 的托管封装。
- TensorRtRuntimeConfig
Wraps TensorRT
IRuntimeConfigfor execution-context creation. 封装 TensorRTIRuntimeConfig，用于创建 execution context。
- TensorRtRuntimeProbeReport
TensorRT runtime staged probe report. TensorRT runtime 分阶段探针报告。
- TensorRtRuntimeProbeStage
Result for one TensorRT runtime probe stage. TensorRT runtime 分阶段探针的单阶段结果。
- TensorRtSerializationConfig
Wraps TensorRT
ISerializationConfigfor engine serialization options. 封装 TensorRTISerializationConfig，用于控制 engine 序列化选项。
- TensorRtSerializedNetworkWithKernelText
Owns a serialized TensorRT engine plan and the optional kernel text emitted by TensorRT 11. 持有 TensorRT 11 生成的序列化 engine plan，以及可选的 kernel text 输出。
- TensorRtSharpInfo
Repository-level metadata for the managed TensorRT assembly. 托管 TensorRT 程序集的仓库级元数据。
- TensorRtTensor
Represents a TensorRT network tensor handle. 表示一个 TensorRT 网络张量句柄。
- TensorRtTensorBindingState
Describes one engine tensor's binding state inside a TensorRT execution context. 描述 TensorRT execution context 中某个 engine tensor 的绑定状态。
- TensorRtTensorInfo
Describes one TensorRT input or output tensor. 描述一个 TensorRT 输入或输出 tensor。
- TensorRtTimingCache
Managed wrapper around a TensorRT timing cache. TensorRT timing cache 的托管封装。
- TensorRtWeights
Immutable TensorRT weights payload for layer creation APIs. 用于 layer 创建 API 的不可变 TensorRT weights 负载。
- TensorRtWeightsInfo
Describes TensorRT weights metadata without exposing the native weights pointer. 描述 TensorRT weights 的元数据，但不暴露原生 weights 指针。
Structs
- TensorRtRefitEntry
Describes a TensorRT layer/role pair that can be refitted or is still missing. 描述一个可 refit 或仍缺失的 TensorRT layer/role 组合。
- TensorRtTimingCacheValue
Represents one TensorRT timing-cache entry value. 表示一个 TensorRT timing cache 条目值。
Enums
- TensorRtActivationType
Represents TensorRT TensorRtActivationType values. 表示 TensorRT TensorRtActivationType 枚举值。
- TensorRtAttentionIoForm
Describes TensorRT attention input/output tensor packing. 描述 TensorRT attention 输入/输出张量排布方式。
- TensorRtAttentionNormalizationOperation
Describes the normalization operation used inside TensorRT attention. 描述 TensorRT attention 内部使用的归一化操作。
- TensorRtBoundingBoxFormat
Describes the coordinate encoding used by a TensorRT NMS layer. 描述 TensorRT NMS 层使用的边界框坐标编码方式。
- TensorRtBuilderFlag
Represents TensorRT TensorRtBuilderFlag values. 表示 TensorRT TensorRtBuilderFlag 枚举值。
- TensorRtBuilderFlags
Represents a TensorRT builder flag bitmask. 表示 TensorRT builder flag 位掩码。
- TensorRtCausalMaskKind
Describes TensorRT attention causal-mask alignment. 描述 TensorRT attention 因果 mask 的对齐方向。
- TensorRtCollectiveOperation
Selects the collective communication primitive for a TensorRT 11 DistCollective layer. 选择 TensorRT 11 DistCollective 层的集合通信原语。
- TensorRtCumulativeOperation
Selects the TensorRT cumulative operation. 选择 TensorRT cumulative 层的累计运算。
- TensorRtDataType
Represents TensorRT TensorRtDataType values. 表示 TensorRT TensorRtDataType 枚举值。
- TensorRtDeviceType
Represents TensorRT TensorRtDeviceType values. 表示 TensorRT TensorRtDeviceType 枚举值。
- TensorRtDistributedReduceOperation
Selects the distributed reduction operation for a TensorRT 11 DistCollective layer. 选择 TensorRT 11 DistCollective 层使用的分布式归约操作。
- TensorRtElementWiseOperation
Represents TensorRT TensorRtElementWiseOperation values. 表示 TensorRT TensorRtElementWiseOperation 枚举值。
- TensorRtEngineCapability
Represents TensorRT TensorRtEngineCapability values. 表示 TensorRT TensorRtEngineCapability 枚举值。
- TensorRtEngineStat
Selects an engine statistic reported by TensorRT. 选择 TensorRT 返回的 engine 统计项。
- TensorRtExecutionContextAllocationStrategy
Selects how TensorRT 11 allocates execution-context device memory. 选择 TensorRT 11 execution context 的设备内存分配策略。
- TensorRtFillOperation
Represents TensorRT TensorRtFillOperation values. 表示 TensorRT TensorRtFillOperation 枚举值。
- TensorRtGatherMode
Selects the TensorRT gather semantics. 选择 TensorRT gather 层的索引语义。
- TensorRtHardwareCompatibilityLevel
Describes TensorRT engine hardware-compatibility requirements. 描述 TensorRT engine 的硬件兼容性要求。
- TensorRtIOMode
Represents TensorRT TensorRtIOMode values. 表示 TensorRT TensorRtIOMode 枚举值。
- TensorRtInterpolationMode
Selects the TensorRT interpolation algorithm for resize and grid-sample style layers. 选择 TensorRT resize / grid-sample 等层使用的插值算法。
- TensorRtKvCacheMode
Describes the TensorRT KV cache update mode. 描述 TensorRT KV cache update 模式。
- TensorRtLayerInformationFormat
Represents TensorRT TensorRtLayerInformationFormat values. 表示 TensorRT TensorRtLayerInformationFormat 枚举值。
- TensorRtLayerType
Represents TensorRT TensorRtLayerType values. 表示 TensorRT TensorRtLayerType 枚举值。
- TensorRtLoopOutputKind
Selects the output semantics of a TensorRT loop output layer. 选择 TensorRT loop output 层的输出语义。
- TensorRtMatrixOperation
Represents TensorRT TensorRtMatrixOperation values. 表示 TensorRT TensorRtMatrixOperation 枚举值。
- TensorRtMemoryPoolType
Represents TensorRT TensorRtMemoryPoolType values. 表示 TensorRT TensorRtMemoryPoolType 枚举值。
- TensorRtMoEActivationType
Describes the activation used by a TensorRT MoE layer. 描述 TensorRT MoE 层使用的激活函数。
- TensorRtNativeDependencySource
Identifies how a native dependency path was discovered by ProbeNativeDependencies(TensorRtApiLine). 标识 ProbeNativeDependencies(TensorRtApiLine) 发现 native 依赖路径的方式。
- TensorRtNetworkDefinitionCreationFlags
Represents TensorRT TensorRtNetworkDefinitionCreationFlags values. 表示 TensorRT TensorRtNetworkDefinitionCreationFlags 枚举值。
- TensorRtOnnxParserFlag
Identifies an individual ONNX parser flag. 标识单个 ONNX parser 标志。
- TensorRtOnnxParserFlags
Represents a bitmask of ONNX parser flags. 表示 ONNX parser 标志位掩码。
- TensorRtOptimizationProfileSelector
Represents TensorRT TensorRtOptimizationProfileSelector values. 表示 TensorRT TensorRtOptimizationProfileSelector 枚举值。
- TensorRtPaddingMode
Represents TensorRT TensorRtPaddingMode values. 表示 TensorRT TensorRtPaddingMode 枚举值。
- TensorRtPluginFieldType
TensorRT plugin field data type reported by plugin creators. TensorRT plugin creator 报告的字段数据类型。
- TensorRtPluginRegistrySource
Identifies the TensorRT registry source used for a plugin creator inventory. 标识 plugin creator inventory 使用的 TensorRT registry 来源。
- TensorRtPoolingType
Represents TensorRT TensorRtPoolingType values. 表示 TensorRT TensorRtPoolingType 枚举值。
- TensorRtPreviewFeature
Selects TensorRT builder preview features. Some values are valid only on specific TensorRT major versions. 选择 TensorRT builder 预览特性；部分取值仅在特定 TensorRT 大版本中有效。
- TensorRtProfilingVerbosity
Represents TensorRT TensorRtProfilingVerbosity values. 表示 TensorRT TensorRtProfilingVerbosity 枚举值。
- TensorRtReduceOperation
Represents TensorRT TensorRtReduceOperation values. 表示 TensorRT TensorRtReduceOperation 枚举值。
- TensorRtResizeCoordinateTransformation
Represents TensorRT TensorRtResizeCoordinateTransformation values. 表示 TensorRT TensorRtResizeCoordinateTransformation 枚举值。
- TensorRtResizeMode
Represents TensorRT TensorRtResizeMode values. 表示 TensorRT TensorRtResizeMode 枚举值。
- TensorRtResizeRoundMode
Represents TensorRT TensorRtResizeRoundMode values. 表示 TensorRT TensorRtResizeRoundMode 枚举值。
- TensorRtResizeSelector
Represents TensorRT TensorRtResizeSelector values. 表示 TensorRT TensorRtResizeSelector 枚举值。
- TensorRtRuntimePlatform
Selects TensorRT 10.x cross-platform runtime target. 选择 TensorRT 10.x 跨平台 runtime 目标。
- TensorRtSampleMode
Controls how TensorRT samples out-of-bounds coordinates. 控制 TensorRT 对越界坐标的采样策略。
- TensorRtScaleMode
Represents TensorRT TensorRtScaleMode values. 表示 TensorRT TensorRtScaleMode 枚举值。
- TensorRtScatterMode
Selects the TensorRT scatter semantics. 选择 TensorRT scatter 层的写入语义。
- TensorRtSerializationFlag
Identifies TensorRT 11 serialization flags. 标识 TensorRT 11 serialization config 的单个标志。
- TensorRtSerializationFlags
Represents TensorRT 11 serialization flags as a bitmask. 以位掩码形式表示 TensorRT 11 serialization flags。
- TensorRtSliceMode
Represents TensorRT TensorRtSliceMode values. 表示 TensorRT TensorRtSliceMode 枚举值。
- TensorRtTacticSources
Represents TensorRT TensorRtTacticSources values. 表示 TensorRT TensorRtTacticSources 枚举值。
- TensorRtTempfileControlFlag
Identifies TensorRT runtime temporary-file control flags. 标识 TensorRT runtime 临时文件控制标志。
- TensorRtTempfileControlFlags
Represents TensorRT runtime temporary-file control flags as a bitmask. 以位掩码形式表示 TensorRT runtime 临时文件控制标志。
- TensorRtTensorFormat
Represents TensorRT TensorRtTensorFormat values. 表示 TensorRT TensorRtTensorFormat 枚举值。
- TensorRtTensorFormats
Represents TensorRT TensorRtTensorFormats values. 表示 TensorRT TensorRtTensorFormats 枚举值。
- TensorRtTensorLocation
Represents TensorRT TensorRtTensorLocation values. 表示 TensorRT TensorRtTensorLocation 枚举值。
- TensorRtTilingOptimizationLevel
Selects TensorRT 11 tiling optimization effort. 选择 TensorRT 11 tiling 优化搜索强度。
- TensorRtTopKOperation
Represents TensorRT TensorRtTopKOperation values. 表示 TensorRT TensorRtTopKOperation 枚举值。
- TensorRtTripLimitKind
Selects how TensorRT limits loop iteration count. 选择 TensorRT 如何限制循环迭代次数。
- TensorRtUnaryOperation
Represents TensorRT TensorRtUnaryOperation values. 表示 TensorRT TensorRtUnaryOperation 枚举值。
- TensorRtWeightsRole
Identifies how TensorRT uses a weight tensor during refitting. 标识 TensorRT 在 refit 过程中如何使用某个权重张量。