TRT - set to floating point fp 32 bits mode INFO: TRT - set to floating point fp 32 bits mode. INFO: Set default type to GPU. TRT Initialization start parsering operation !!! ---------------------------------------------------------------- Input filename: OnnxMultiDynamicInputsTestDynamicModel.onnx ONNX IR version: 0.0.6 Opset version: 9 Producer name: pytorch Producer version: 1.7 Domain: Model version: 0 Doc string: ---------------------------------------------------------------- VERBOSE: Registered plugin creator - ::GridAnchor_TRT version 1 VERBOSE: Registered plugin creator - ::NMS_TRT version 1 VERBOSE: Registered plugin creator - ::Reorg_TRT version 1 VERBOSE: Registered plugin creator - ::Region_TRT version 1 VERBOSE: Registered plugin creator - ::Clip_TRT version 1 VERBOSE: Registered plugin creator - ::LReLU_TRT version 1 VERBOSE: Registered plugin creator - ::PriorBox_TRT version 1 VERBOSE: Registered plugin creator - ::Normalize_TRT version 1 VERBOSE: Registered plugin creator - ::RPROI_TRT version 1 VERBOSE: Registered plugin creator - ::BatchedNMS_TRT version 1 VERBOSE: Registered plugin creator - ::BatchedNMSDynamic_TRT version 1 VERBOSE: Registered plugin creator - ::FlattenConcat_TRT version 1 VERBOSE: Registered plugin creator - ::CropAndResize version 1 VERBOSE: Registered plugin creator - ::DetectionLayer_TRT version 1 VERBOSE: Registered plugin creator - ::Proposal version 1 VERBOSE: Registered plugin creator - ::ProposalLayer_TRT version 1 VERBOSE: Registered plugin creator - ::PyramidROIAlign_TRT version 1 VERBOSE: Registered plugin creator - ::ResizeNearest_TRT version 1 VERBOSE: Registered plugin creator - ::Split version 1 VERBOSE: Registered plugin creator - ::SpecialSlice_TRT version 1 VERBOSE: Registered plugin creator - ::InstanceNormalization_TRT version 1 VERBOSE: ModelImporter.cpp:202: Adding network input: my_input1 with dtype: float32, dimensions: (-1, -1, -1, -1) VERBOSE: ImporterContext.hpp:120: Registering tensor: my_input1 for ONNX tensor: my_input1 VERBOSE: ModelImporter.cpp:202: Adding network input: my_input2 with dtype: float32, dimensions: (-1, -1, -1, -1) VERBOSE: ImporterContext.hpp:120: Registering tensor: my_input2 for ONNX tensor: my_input2 VERBOSE: ModelImporter.cpp:90: Importing initializer: conv1.bias VERBOSE: ModelImporter.cpp:90: Importing initializer: conv1.weight VERBOSE: ModelImporter.cpp:103: Parsing node: Conv_0 [Conv] VERBOSE: ModelImporter.cpp:119: Searching for input: my_input1 VERBOSE: ModelImporter.cpp:119: Searching for input: conv1.weight VERBOSE: ModelImporter.cpp:119: Searching for input: conv1.bias VERBOSE: ModelImporter.cpp:125: Conv_0 [Conv] inputs: [my_input1 -> (-1, -1, -1, -1)], [conv1.weight -> (32, 32, 3, 3)], [conv1.bias -> (32)], VERBOSE: builtin_op_importers.cpp:448: Convolution input dimensions: (-1, -1, -1, -1) VERBOSE: ImporterContext.hpp:154: Registering layer: Conv_0 for ONNX node: Conv_0 VERBOSE: builtin_op_importers.cpp:537: Using kernel: (3, 3), strides: (1, 1), prepadding: (1, 1), postpadding: (1, 1), dilations: (1, 1), numOutputs: 32 VERBOSE: builtin_op_importers.cpp:538: Convolution output dimensions: (-1, 32, -1, -1) VERBOSE: ImporterContext.hpp:120: Registering tensor: 11 for ONNX tensor: 11 VERBOSE: ModelImporter.cpp:179: Conv_0 [Conv] outputs: [11 -> (-1, 32, -1, -1)], VERBOSE: ModelImporter.cpp:103: Parsing node: MaxPool_1 [MaxPool] VERBOSE: ModelImporter.cpp:119: Searching for input: my_input2 VERBOSE: ModelImporter.cpp:125: MaxPool_1 [MaxPool] inputs: [my_input2 -> (-1, -1, -1, -1)], VERBOSE: ImporterContext.hpp:154: Registering layer: MaxPool_1 for ONNX node: MaxPool_1 ERROR: MaxPool_1: at least 5 dimensions are required for input. ERROR: MaxPool_1: at least 5 dimensions are required for input. VERBOSE: ImporterContext.hpp:120: Registering tensor: 12 for ONNX tensor: 12 VERBOSE: ModelImporter.cpp:179: MaxPool_1 [MaxPool] outputs: [12 -> ()], VERBOSE: ModelImporter.cpp:103: Parsing node: Add_2 [Add] VERBOSE: ModelImporter.cpp:119: Searching for input: 11 VERBOSE: ModelImporter.cpp:119: Searching for input: 12 ERROR: MaxPool_1: at least 5 dimensions are required for input. VERBOSE: ModelImporter.cpp:125: Add_2 [Add] inputs: [11 -> (-1, 32, -1, -1)], [12 -> ()], ERROR: MaxPool_1: at least 5 dimensions are required for input. ERROR: MaxPool_1: at least 5 dimensions are required for input. ERROR: MaxPool_1: at least 5 dimensions are required for input. WARNING: onnx2trt_utils.cpp:220: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32. ERROR: MaxPool_1: at least 5 dimensions are required for input. ERROR: MaxPool_1: at least 5 dimensions are required for input. ERROR: MaxPool_1: at least 5 dimensions are required for input. ERROR: onnx2trt_utils.cpp:680 In function elementwiseHelper: [8] Assertion failed: tensor_ptr->getDimensions().nbDims == maxNbDims && "Failed to broadcast tensors elementwise!" Error code - 8 Error description - Assertion failed: tensor_ptr->getDimensions().nbDims == maxNbDims && "Failed to broadcast tensors elementwise!" Error source file - onnx2trt_utils.cpp Error source line - 680 Error function - elementwiseHelper Error node - -1 ERROR: Fail to parse Runtime Exception - Onnx parsering operation error!!!!!!