[Help]: Unable to build and use the 'tensorrt_yolo' package correctly

[dyhisbig]

较为详细的教程.docx 希望作者可以将大家遇到的问题整理成一个reamde文件，这样后续找到者这的人会少一些弯路

[laugh12321]

3.0 的推理也是GPU推理，只不过没有cudaGraph加速。而且你图上贴的是4.0的报错，我怀疑你没有卸载之前安装的tensorrt_yolo再pip install tensorrt_yolo安装3.0版本，3.0版本是不可能出现找不到指定模块的错误的。

[laugh12321]

https://github.com/user-attachments/assets/dcb0e078-c736-49cd-84ab-5393cbf78c20

@dyhisbig 这是在一个新环境上的TensorRT-YOLO 使用教程，其中的CUDA、cuDNN、TensorRT、Xmake的配置我这里不去赘述。你看看你和这个流程一样吗？都是README上面的指令一个都没变。

[dyhisbig]

TensorRT-YOLO.Tutorial.mp4

@dyhisbig 这是在一个新环境上的TensorRT-YOLO 使用教程，其中的CUDA、cuDNN、TensorRT、Xmake的配置我这里不去赘述。你看看你和这个流程一样吗？都是README上面的指令一个都没变。

我按照视频里，重新建了个环境，前边一切顺利，和视频里一样，最后执行推理trtyolo infer这个命令就是上述错误

[laugh12321]

@dyhisbig 你的Python版本和xmake.lua要求的一样吗？xmake.lua中默认是3.10

[dyhisbig]

我在试试，不挂梯子报错那个图片传不上来

[laugh12321]

@dyhisbig 根据你上面的文档，我怀疑你编译的是3.10的包，然后又给3.11环境去安装，那么同样是会出现找不到指定模块的错误的。

[dyhisbig]

换了个3.10的python环境执infer命令出现了新的问题，刚才是3.8的环境，是导包的问题， Traceback (most recent call last): File "C:\dyh\annconda\envs\py10\lib\runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, return callback(*args, **kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 161, in infer model = DeployDet(engine) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 45, in init__ self._model = C.detection.DeployDet(engine_file, use_cuda_memory, device) File "C:\dyh\annconda\envs\py10\lib\runpy.py", line 86, in _run_code exec(code, run_globals) File "C:\dyh\annconda\envs\py10\Scripts\trtyolo.exe__main.py", line 7, in File "C:\dyh\annconda\envs\py10\lib\site-packages\rich_click\rich_command.py", line 367, in call return super().call(*args, **kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1157, in call return self.main(*args, kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\rich_click\rich_command.py", line 152, in main rv = self.invoke(ctx) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1688, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1434, in invoke return ctx.invoke(self.callback, ctx.params) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 783, in invoke return callback(*args, **kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 161, in infer model = DeployDet(engine) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 45, in init self._model = C.detection.DeployDet(engine_file, use_cuda_memory, device) RuntimeError: Failed to construct engine context.

[dyhisbig]

python --version Python 3.10.14

[dyhisbig]

再次执行那行命令，中间多了个错误 ERROR: 1: [stdArchiveReader.cpp::nvinfer1::rt::StdArchiveReader::StdArchiveReaderInitCommon::46] Error Code 1: Serialization (Serialization assertion stdVersionRead == serializationVersion failed.Version tag does not match. Note: Current Version: 236, Serialized Engine Version: 238) 原地起飞

[laugh12321]

换了个3.10的python环境执infer命令出现了新的问题，刚才是3.8的环境，是导包的问题， Traceback (most recent call last): File "C:\dyh\annconda\envs\py10\lib\runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, return __callback(*args, kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 161, in infer model = DeployDet(engine) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 45, in init self._model = C.detection.DeployDet(engine_file, use_cuda_memory, device) File "C:\dyh\annconda\envs\py10\lib\runpy.py", line 86, in _run_code exec(code, run_globals) File "C:\dyh\annconda\envs\py10\Scripts\trtyolo.exemain.py", line 7, in File "C:\dyh\annconda\envs\py10\lib\site-packages\rich_click\rich_command.py", line 367, in call return super().call(*args, kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1157, in call* return self.main(args, kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\rich_click\rich_command.py", line 152, in main rv = self.invoke(ctx) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1688, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1434, in invoke return ctx.invoke(self.callback, *ctx.params) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 783, in invoke return __callback(args, kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 161, in infer model = DeployDet(engine) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 45, in init** self._model = C.detection.DeployDet(engine_file, use_cuda_memory, device) RuntimeError: Failed to construct engine context.

这一行出发的报错 检查一下模型路径是否正确，或者重新用trtexec导出模型并给出正确的路径。

[laugh12321]

再次执行那行命令，中间多了个错误 ERROR: 1: [stdArchiveReader.cpp::nvinfer1::rt::StdArchiveReader::StdArchiveReaderInitCommon::46] Error Code 1: Serialization (Serialization assertion stdVersionRead == serializationVersion failed.Version tag does not match. Note: Current Version: 236, Serialized Engine Version: 238) 原地起飞

你是不是有多个版本的TensorRT，错误意思是编译的TensorRT版本与Engine的版本不一致。

[dyhisbig]

[I] Successfully found necessary library paths: { "cudart": "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1\bin", "nvinfer": "E:\TensorRT-10.2.0.19.Windows.win10.cuda-12.5\TensorRT-10.2.0.19\bin", "cudnn": "D:\anaconda3\envs\oytorch\Lib\site-packages\torch\lib" } File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1434, in invoke return ctx.invoke(self.callback, *ctx.params) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 783, in invoke return __callback(args, **kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 151, in infer from .infer import CpuTimer, DeployCGDet, DeployDet, GpuTimer, generate_labels_with_colors, visualize_detections File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer__init__.py", line 1, in from .detection import Box, DeployCGDet, DeployDet, DetectionResult File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 27, in from .. import c_lib_wrap as C File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\c_lib_wrap.py", line 150, in raise RuntimeError(f"Deploy initialization failed! Error: {e}") RuntimeError: Deploy initialization failed! Error: DLL load failed while importing pydeploy: 找不到指定的模块。 我现在用etnsorrt10.2编译，deploy，导出engine，好像又回到了起点，哈哈哈

[laugh12321]

@dyhisbig 首先前面报错[Version tag does not match. Note: Current Version: 236, Serialized Engine Version: 238]证明之前那个是打包步骤正确的，所以没有出现[找不到指定的模块]这个错误，只不过是你打包的TensorRT版本与让它推理的engine版本不一致而已，所以你只需要使用打包版本的trtexec重新生成engine即可。

其次，对于你这次出现这个问题。有以下几种原因

• 可能是编译Python的版本与安装环境的Python版本不一致。
• 没有重新执行完整的流程（编译加打包），只是重新打包。
• 没有删除lib与tensorrt_yolo/libs下的编译文件后在重新执行完整流程。
• cudnn 的路径不应该在"D:\anaconda3\envs\oytorch\Lib\site-packages\torch\lib"

最后，因为你环境配置出现差错导致的tensorrt_yolo无法正常使用的问题，这里将不再解答。建议参考【Windows 开发环境配置——NVIDIA 篇】CUDA、cuDNN、TensorRT 三件套安装 正确的配置相关环境。

以下是正确编译并使用tensorr_yolo包的前提：

• 编译Deploy模块时的Python版本要与xmake.lua中设置的Python版本以及使用环境的Python版本一致；
• 正确的安装CUDA、cuDNN、TensorRT 三件套安装并配置环境变量；
• 尽可能保证没有多个版本的CUDA、cuDNN、TensorRT以防查找到错误的dll路径；
• 推理时要保证tensorrt_yolo构建使用的TensorRT版本与生成engine的版本要一致。

在满足以上基础前提且操作顺序与教程一致的情况下，若仍出现问题再来提问。

[dyhisbig]

[I] Successfully found necessary library paths: { "cudart": "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.5\bin", "cudnn": "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.5\bin", "nvinfer": "C:\Program Files\NVIDIA GPU Computing Toolkit\TensorRT\v8.6.1.6\bin" } Traceback (most recent call last): File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\c_lib_wrap.py", line 148, in from .libs.pydeploy import * rv = self.invoke(ctx) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1688, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1434, in invoke return ctx.invoke(self.callback, *ctx.params) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 783, in invoke return __callback(args, **kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 151, in infer from .infer import CpuTimer, DeployCGDet, DeployDet, GpuTimer, generate_labels_with_colors, visualize_detections File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer__init__.py", line 1, in from .detection import Box, DeployCGDet, DeployDet, DetectionResult File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 27, in from .. import c_lib_wrap as C File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\c_lib_wrap.py", line 150, in raise RuntimeError(f"Deploy initialization failed! Error: {e}") RuntimeError: Deploy initialization failed! Error: DLL load failed while importing pydeploy: 找不到指定的模块。 1、编译环境安装环境都py3.10 2、删掉文件重新执行打包、编译，都试过了，cuda cudnn trt环境变量都配置了，trt换成了8.8.1.6 3、cudnn路径现在也识别队对了 4、构建、生成engine的trt都是8.6.1.6

[dyhisbig]

[I] Successfully found necessary library paths: { "cudart": "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.5\bin", "cudnn": "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.5\bin", "nvinfer": "C:\Program Files\NVIDIA GPU Computing Toolkit\TensorRT\v8.6.1.6\bin" } Traceback (most recent call last): File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\c_lib_wrap.py", line 148, in from .libs.pydeploy import * rv = self.invoke(ctx) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1688, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 1434, in invoke return ctx.invoke(self.callback, *ctx.params) File "C:\dyh\annconda\envs\py10\lib\site-packages\click\core.py", line 783, in invoke return __callback(args, **kwargs) File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\cli.py", line 151, in infer from .infer import CpuTimer, DeployCGDet, DeployDet, GpuTimer, generate_labels_with_colors, visualize_detections File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer__init__.py", line 1, in from .detection import Box, DeployCGDet, DeployDet, DetectionResult File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\infer\detection.py", line 27, in from .. import c_lib_wrap as C File "C:\dyh\annconda\envs\py10\lib\site-packages\tensorrt_yolo\c_lib_wrap.py", line 150, in raise RuntimeError(f"Deploy initialization failed! Error: {e}") RuntimeError: Deploy initialization failed! Error: DLL load failed while importing pydeploy: 找不到指定的模块。 1、编译环境安装环境都py3.10 2、删掉文件重新执行打包、编译，都试过了，cuda cudnn trt环境变量都配置了，trt换成了8.8.1.6 3、cudnn路径现在也识别队对了 4、构建、生成engine的trt都是8.6.1.6

[laugh12321]

邮箱私信我，发你Todesk我帮你解决。

[dyhisbig]

解决了，作者亲自上阵，虽然代码不一定对我有用，但是捣鼓两天不死心，是cuda里边的放了别的版本的cudnn，但是也很奇怪，我是下午重新装了的cuda，待我在研究研究

[laugh12321]

解决了，作者亲自上阵，虽然代码不一定对我有用，但是捣鼓两天不死心，是cuda里边的放了别的版本的cudnn，但是也很奇怪，我是下午重新装了的cuda，待我在研究研究

CUDA卸载时，不会删除手动放置在C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA下的文件，所以你重装后，之前放置的cuDNN还是存在的。