pointpillars模型导出失败，缺少pointpillars.pdiparams.info和pointpillars..pdmodel文件，尝试使用.json和.pdiparams进行部署推理时报错。

[yyan-w]

1. 在通过tools/export.py导出下载的官方推理模型model.pdparams时，只生成了pointpillars.json和pointpillars.dpiparams两个文件。和官方readme.md中所言”保存导出模型的路径，save_dir下将会生成三个文件：pointpillars.pdiparams 、pointpillars.pdiparams.info和pointpillars.pdmodel“不同。

2. 由于未能成功导出pointpillars.pdmodel，于是尝试使用目前导出的pointpillars.json代替pointpillars.pdmodel作为--model_file 参数进行基于python的部署推理时，出现报错。以下是我的代码及报错信息：
   推理命令：
   python infer.py \ --model_file /home/user/Paddle3D/output/pointpillars/pointpillars.json \ --params_file /home/user/Paddle3D/output/pointpillars/pointpillars.pdiparams \ --lidar_file /home/share/PCD-kitti/kitti_dataset_root/training/velodyne/000008.bin \ --point_cloud_range 0 -39.68 -3 69.12 39.68 1 \ --voxel_size .16 .16 4 \ --max_points_in_voxel 32 \ --max_voxel_num 40000
   报错如下：
   `I0314 14:58:52.662317 3012545 analysis_predictor.cc:1142] ======= pir optimization completed =======
   I0314 14:58:53.447067 3012545 pir_interpreter.cc:1607] pir interpreter is running by trace mode ...
   W0314 14:58:53.732611 3012545 pir_interpreter.cc:1996] Instruction OP id: 41, Ir OP id: 3723, pd_op.stack raises an EnforceNotMet exception common::enforce::EnforceNotMet
   W0314 14:58:53.732690 3012545 pir_interpreter.cc:2002] Instruction OP id: 20, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet
   W0314 14:58:53.732723 3012545 pir_interpreter.cc:2002] Instruction OP id: 136, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet
   Traceback (most recent call last):
   File "infer.py", line 281, in
   main(args)
   File "infer.py", line 273, in main
   box3d_lidar, label_preds, scores = run(predictor, voxels, coords,
   File "infer.py", line 250, in run
   predictor.run()
   ValueError: In user code:

   InvalidArgumentError: The type of data we are trying to retrieve (int64) does not match the type of data (int32) currently contained in the container.
   [Hint: Expected dtype() == phi::CppTypeToDataType::Type(), but received dtype():7 != phi::CppTypeToDataType::Type():9.] (at ../paddle/phi/core/dense_tensor.cc:160)
   [operator < pd_op.if > error]
   `

我的环境主要配置如下：
paddlepaddle-gpu==3.0.0rc1
paddle3d==0.0.0
cuda=11.8
请问问题是否源于模型导出？应该如何解决呢？

[1357810la]

1. 在通过tools/export.py导出下载的官方推理模型model.pdparams时，只生成了pointpillars.json和pointpillars.dpiparams两个文件。和官方readme.md中所言”保存导出模型的路径，save_dir下将会生成三个文件：pointpillars.pdiparams 、pointpillars.pdiparams.info和pointpillars.pdmodel“不同。
2. 由于未能成功导出pointpillars.pdmodel，于是尝试使用目前导出的pointpillars.json代替pointpillars.pdmodel作为--model_file 参数进行基于python的部署推理时，出现报错。以下是我的代码及报错信息：
   推理命令：
   python infer.py \ --model_file /home/user/Paddle3D/output/pointpillars/pointpillars.json \ --params_file /home/user/Paddle3D/output/pointpillars/pointpillars.pdiparams \ --lidar_file /home/share/PCD-kitti/kitti_dataset_root/training/velodyne/000008.bin \ --point_cloud_range 0 -39.68 -3 69.12 39.68 1 \ --voxel_size .16 .16 4 \ --max_points_in_voxel 32 \ --max_voxel_num 40000
   报错如下：
   I0314 14:58:52.662317 3012545 analysis_predictor.cc:1142] ======= pir optimization completed ======= I0314 14:58:53.447067 3012545 pir_interpreter.cc:1607] pir interpreter is running by trace mode ... W0314 14:58:53.732611 3012545 pir_interpreter.cc:1996] Instruction OP id: 41, Ir OP id: 3723, pd_op.stack raises an EnforceNotMet exception common::enforce::EnforceNotMet W0314 14:58:53.732690 3012545 pir_interpreter.cc:2002] Instruction OP id: 20, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet W0314 14:58:53.732723 3012545 pir_interpreter.cc:2002] Instruction OP id: 136, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet Traceback (most recent call last): File "infer.py", line 281, in main(args) File "infer.py", line 273, in main box3d_lidar, label_preds, scores = run(predictor, voxels, coords, File "infer.py", line 250, in run predictor.run() ValueError: In user code: InvalidArgumentError: The type of data we are trying to retrieve (int64) does not match the type of data (int32) currently contained in the container. [Hint: Expected dtype() == phi::CppTypeToDataType::Type(), but received dtype():7 != phi::CppTypeToDataType::Type():9.] (at ../paddle/phi/core/dense_tensor.cc:160) [operator < pd_op.if > error]

我的环境主要配置如下： paddlepaddle-gpu==3.0.0rc1 paddle3d==0.0.0 cuda=11.8 请问问题是否源于模型导出？应该如何解决呢？

你好，请问解决了吗，我导出时也出现缺少.pdmodel和.info文件

[yyan-w]

你好，我也还没有解决，而且社区里也没有任何回复

…

---原始邮件--- 发件人: ***@***.***&gt; 发送时间: 2025年4月17日(周四) 上午10:29 收件人: ***@***.***&gt;; 抄送: ***@***.******@***.***&gt;; 主题: Re: [PaddlePaddle/Paddle3D] pointpillars模型导出失败，缺少pointpillars.pdiparams.info和pointpillars..pdmodel文件，尝试使用.json和.pdiparams进行部署推理时报错。 (Issue #494) 在通过tools/export.py导出下载的官方推理模型model.pdparams时，只生成了pointpillars.json和pointpillars.dpiparams两个文件。和官方readme.md中所言”保存导出模型的路径，save_dir下将会生成三个文件：pointpillars.pdiparams 、pointpillars.pdiparams.info和pointpillars.pdmodel“不同。 由于未能成功导出pointpillars.pdmodel，于是尝试使用目前导出的pointpillars.json代替pointpillars.pdmodel作为--model_file 参数进行基于python的部署推理时，出现报错。以下是我的代码及报错信息： 推理命令： python infer.py \ --model_file /home/user/Paddle3D/output/pointpillars/pointpillars.json \ --params_file /home/user/Paddle3D/output/pointpillars/pointpillars.pdiparams \ --lidar_file /home/share/PCD-kitti/kitti_dataset_root/training/velodyne/000008.bin \ --point_cloud_range 0 -39.68 -3 69.12 39.68 1 \ --voxel_size .16 .16 4 \ --max_points_in_voxel 32 \ --max_voxel_num 40000 报错如下： I0314 14:58:52.662317 3012545 analysis_predictor.cc:1142] ======= pir optimization completed ======= I0314 14:58:53.447067 3012545 pir_interpreter.cc:1607] pir interpreter is running by trace mode ... W0314 14:58:53.732611 3012545 pir_interpreter.cc:1996] Instruction OP id: 41, Ir OP id: 3723, pd_op.stack raises an EnforceNotMet exception common::enforce::EnforceNotMet W0314 14:58:53.732690 3012545 pir_interpreter.cc:2002] Instruction OP id: 20, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet W0314 14:58:53.732723 3012545 pir_interpreter.cc:2002] Instruction OP id: 136, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet Traceback (most recent call last): File "infer.py", line 281, in main(args) File "infer.py", line 273, in main box3d_lidar, label_preds, scores = run(predictor, voxels, coords, File "infer.py", line 250, in run predictor.run() ValueError: In user code: InvalidArgumentError: The type of data we are trying to retrieve (int64) does not match the type of data (int32) currently contained in the container. [Hint: Expected dtype() == phi::CppTypeToDataType::Type(), but received dtype():7 != phi::CppTypeToDataType::Type():9.] (at ../paddle/phi/core/dense_tensor.cc:160) [operator < pd_op.if &gt; error] 我的环境主要配置如下： paddlepaddle-gpu==3.0.0rc1 paddle3d==0.0.0 cuda=11.8 请问问题是否源于模型导出？应该如何解决呢？ 你好，请问解决了吗，我导出时也出现缺少.pdmodel和.info文件 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt; 1357810la left a comment (PaddlePaddle/Paddle3D#494) 在通过tools/export.py导出下载的官方推理模型model.pdparams时，只生成了pointpillars.json和pointpillars.dpiparams两个文件。和官方readme.md中所言”保存导出模型的路径，save_dir下将会生成三个文件：pointpillars.pdiparams 、pointpillars.pdiparams.info和pointpillars.pdmodel“不同。 由于未能成功导出pointpillars.pdmodel，于是尝试使用目前导出的pointpillars.json代替pointpillars.pdmodel作为--model_file 参数进行基于python的部署推理时，出现报错。以下是我的代码及报错信息： 推理命令： python infer.py \ --model_file /home/user/Paddle3D/output/pointpillars/pointpillars.json \ --params_file /home/user/Paddle3D/output/pointpillars/pointpillars.pdiparams \ --lidar_file /home/share/PCD-kitti/kitti_dataset_root/training/velodyne/000008.bin \ --point_cloud_range 0 -39.68 -3 69.12 39.68 1 \ --voxel_size .16 .16 4 \ --max_points_in_voxel 32 \ --max_voxel_num 40000 报错如下： I0314 14:58:52.662317 3012545 analysis_predictor.cc:1142] ======= pir optimization completed ======= I0314 14:58:53.447067 3012545 pir_interpreter.cc:1607] pir interpreter is running by trace mode ... W0314 14:58:53.732611 3012545 pir_interpreter.cc:1996] Instruction OP id: 41, Ir OP id: 3723, pd_op.stack raises an EnforceNotMet exception common::enforce::EnforceNotMet W0314 14:58:53.732690 3012545 pir_interpreter.cc:2002] Instruction OP id: 20, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet W0314 14:58:53.732723 3012545 pir_interpreter.cc:2002] Instruction OP id: 136, Ir OP id is null, if_instruction raises an EnforceNotMet exception common::enforce::EnforceNotMet Traceback (most recent call last): File "infer.py", line 281, in main(args) File "infer.py", line 273, in main box3d_lidar, label_preds, scores = run(predictor, voxels, coords, File "infer.py", line 250, in run predictor.run() ValueError: In user code: InvalidArgumentError: The type of data we are trying to retrieve (int64) does not match the type of data (int32) currently contained in the container. [Hint: Expected dtype() == phi::CppTypeToDataType::Type(), but received dtype():7 != phi::CppTypeToDataType::Type():9.] (at ../paddle/phi/core/dense_tensor.cc:160) [operator < pd_op.if &gt; error] 我的环境主要配置如下： paddlepaddle-gpu==3.0.0rc1 paddle3d==0.0.0 cuda=11.8 请问问题是否源于模型导出？应该如何解决呢？ 你好，请问解决了吗，我导出时也出现缺少.pdmodel和.info文件 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt;

[xukkko]

建议安装低版本的paddle，详见：https://blog.csdn.net/gitblog_01410/article/details/150635243