search

NVIDIA-AI-IOT / Lidar_AI_Solution

A project demonstrating Lidar related AI solutions, including three GPU accelerated Lidar/camera DL networks (PointPillars, CenterPoint, BEVFusion) and the related libs (cuPCL, 3D SparseConvolution, YUV2RGB, cuOSD,).
Other
1.26k stars 220 forks source link

this error when I export epoch.pth to onnx with the export_neck_head.py, but the filename.onnx can work, please help me #160

Open HULoaf opened 1 year ago

HULoaf commented 1 year ago

Warning: Constant folding - Only steps=1 can be constant folded for opset >= 10 onnx::Slice op. Constant folding not applied. Warning: Constant folding - Only steps=1 can be constant folded for opset >= 10 onnx::Slice op. Constant folding not applied. [PASS] Export ONNX done. /data/chuxikai/anaconda3/envs/nvi_cen/lib/python3.8/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:2157.) return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined] Traceback (most recent call last): File "tool/export_neck_head.py", line 299, in main(args) File "tool/export_neck_head.py", line 280, in main np.testing.assert_almost_equal(box3d_lidar, gt_output_ori['box3d_lidar'].cpu().numpy(), decimal=3) File "/data/chuxikai/anaconda3/envs/nvi_cen/lib/python3.8/site-packages/numpy/testing/_private/utils.py", line 581, in assert_almost_equal return assert_array_almost_equal(actual, desired, decimal, err_msg) File "/data/chuxikai/anaconda3/envs/nvi_cen/lib/python3.8/site-packages/numpy/testing/_private/utils.py", line 1044, in assert_array_almost_equal assert_array_compare(compare, x, y, err_msg=err_msg, verbose=verbose, File "/data/chuxikai/anaconda3/envs/nvi_cen/lib/python3.8/site-packages/numpy/testing/_private/utils.py", line 842, in assert_array_compare raise AssertionError(msg) AssertionError: Arrays are not almost equal to 3 decimals

Mismatched elements: 37 / 1683 (2.2%) Max absolute difference: 38.031326 Max relative difference: 8.296275 x: array([[ 3.433e-01, 3.039e+01, 5.712e-01, ..., 3.006e-01, 8.440e+00, -3.089e+00], [-9.198e+00, -1.790e+01, -1.655e+00, ..., -1.204e-01, -4.462e+00,... y: array([[ 3.433e-01, 3.039e+01, 5.712e-01, ..., 3.004e-01, 8.440e+00, -3.089e+00], [-9.198e+00, -1.790e+01, -1.655e+00, ..., -1.203e-01, -4.461e+00,...

hopef commented 1 year ago

You can comment np.assert_almost_equal function to avoid this error.

HULoaf commented 1 year ago

You can comment np.assert_almost_equal function to avoid this error.

Thanks, it works! By the way, i want to train my custom dataset with only lidar, i have tried for long time, but i filed. Can you give me some advice? Tears of gratitude

HULoaf commented 1 year ago

You can comment np.assert_almost_equal function to avoid this error.

Besides, I have trained my custom dataset with only lidar on MMdetection3d, but I failed to export onnx from the pth i trained with using your export tools including export-scn.py and export_neck_head.py, could you give me some advice, appreciate!

hopef commented 1 year ago

I'm pretty sure it's a difficult process, which means you need to learn the exported code and more knowledge of the model. Thanks!