编译fasttrt时报错,tensorrt版本8.2.1.8

[JianJingGao]

支持trt 8的部署吗

[JianJingGao]

修改好编译成功,make install之后没有出fasttrt... -- Install configuration: "Release" -- Installing: /home/gtt/work/git/fast-reid-master/projects/FastRT/libs/FastRTEngine/libFastRTEngine.so.0.0.5 -- Installing: /home/gtt/work/git/fast-reid-master/projects/FastRT/libs/FastRTEngine/libFastRTEngine.so.0 -- Set runtime path of "/home/gtt/work/git/fast-reid-master/projects/FastRT/libs/FastRTEngine/libFastRTEngine.so.0.0.5" to "" -- Installing: /home/gtt/work/git/fast-reid-master/projects/FastRT/libs/FastRTEngine/libFastRTEngine.so

[JianJingGao]

[ModelConfig]: weights_path: ./vehicle_r50.wts max_batch_size: 4 input_h: 384 input_w: 128 output_size: 2048 device_id: 0

[FastreidConfig]: backbone: r50 head: EmbeddingHead pooling: gempoolP last_stride: 1 with_ibna: 1 with_nl: 1 embedding_dim: 0

[Serializling Engine]

[Non_local] inter_channels: 1 [05/28/2024-16:51:25] [E] [TRT] 3: (Unnamed Layer 132) [Convolution]:kernel weights has count 0 but 512 was expected [05/28/2024-16:51:25] [E] [TRT] 4: (Unnamed Layer 132) [Convolution]: count of 0 weights in kernel, but kernel dimensions (1,1) with 512 input channels, 1 output channels and 1 groups were specified. Expected Weights count is 512 11 1 / 1 = 512 [05/28/2024-16:51:25] [E] [TRT] 4: [convolutionNode.cpp::computeOutputExtents::39] Error Code 4: Internal Error ((Unnamed Layer 132) [Convolution]: number of kernel weights does not match tensor dimensions) [05/28/2024-16:51:25] [E] [TRT] 3: (Unnamed Layer 132) [Convolution]:kernel weights has count 0 but 512 was expected [05/28/2024-16:51:25] [E] [TRT] 4: (Unnamed Layer 132) [Convolution]: count of 0 weights in kernel, but kernel dimensions (1,1) with 512 input channels, 1 output channels and 1 groups were specified. Expected Weights count is 512 11 1 / 1 = 512 [05/28/2024-16:51:25] [E] [TRT] 4: [convolutionNode.cpp::computeOutputExtents::39] Error Code 4: Internal Error ((Unnamed Layer 132) [Convolution]: number of kernel weights does not match tensor dimensions) [05/28/2024-16:51:25] [E] [TRT] 3: (Unnamed Layer 132) [Convolution]:kernel weights has count 0 but 512 was expected [05/28/2024-16:51:25] [E] [TRT] 4: (Unnamed Layer 132) [Convolution]: count of 0 weights in kernel, but kernel dimensions (1,1) with 512 input channels, 1 output channels and 1 groups were specified. Expected Weights count is 512 11 1 / 1 = 512 [05/28/2024-16:51:25] [E] [TRT] 4: [convolutionNode.cpp::computeOutputExtents::39] Error Code 4: Internal Error ((Unnamed Layer 132) [Convolution]: number of kernel weights does not match tensor dimensions) [05/28/2024-16:51:25] [E] [TRT] 3: (Unnamed Layer 132) [Convolution]:kernel weights has count 0 but 512 was expected [05/28/2024-16:51:25] [E] [TRT] 4: (Unnamed Layer 132) [Convolution]: count of 0 weights in kernel, but kernel dimensions (1,1) with 512 input channels, 1 output channels and 1 groups were specified. Expected Weights count is 512 11 1 / 1 = 512 [05/28/2024-16:51:25] [E] [TRT] 4: [convolutionNode.cpp::computeOutputExtents::39] Error Code 4: Internal Error ((Unnamed Layer 132) [Convolution]: number of kernel weights does not match tensor dimensions) [05/28/2024-16:51:25] [E] [TRT] 3: (Unnamed Layer 132) [Convolution]:kernel weights has count 0 but 512 was expected [05/28/2024-16:51:25] [E] [TRT] 4: (Unnamed Layer 132) [Convolution]: count of 0 weights in kernel, but kernel dimensions (1,1) with 512 input channels, 1 output channels and 1 groups were specified. Expected Weights count is 512 11 1 / 1 = 512 [05/28/2024-16:51:25] [E] [TRT] 4: [convolutionNode.cpp::computeOutputExtents::39] Error Code 4: Internal Error ((Unnamed Layer 132) [Convolution]: number of kernel weights does not match tensor dimensions) [05/28/2024-16:51:25] [E] [TRT] 3: [network.cpp::addConstant::1193] Error Code 3: API Usage Error (Parameter check failed at: optimizer/api/network.cpp::addConstant::1193, condition: !weights.values == !weights.count ) 段错误 (核心已转储) 编译出之后wts转engine报错 是因为trt8.0 API原因吗?

[JianJingGao]

改称trt7.2.3没问题了 但是导出vehicle部分出错,可以给点提示吗,手册上没找到这个结构的example

[JianJingGao]

搞定了 对照py的demo.py的结构跑通了