RTX 3090 not compatible

I’ve set up the environment according to the instructions. But when I train the model on my dataset with this script: python main.py --num_classes 5 --num_points 200 --model pointMLP --workers 1 An error occurs:

Traceback (most recent call last):
  File "main.py", line 273, in <module>
    main()
  File "main.py", line 84, in main
    net = net.to(device)
  File "/home/ccbien/.conda/envs/point-mlp/lib/python3.7/site-packages/torch/nn/modules/module.py", line 899, in to
    return self._apply(convert)
  File "/home/ccbien/.conda/envs/point-mlp/lib/python3.7/site-packages/torch/nn/modules/module.py", line 570, in _apply
    module._apply(fn)
  File "/home/ccbien/.conda/envs/point-mlp/lib/python3.7/site-packages/torch/nn/modules/module.py", line 570, in _apply
    module._apply(fn)
  File "/home/ccbien/.conda/envs/point-mlp/lib/python3.7/site-packages/torch/nn/modules/module.py", line 570, in _apply
    module._apply(fn)
  File "/home/ccbien/.conda/envs/point-mlp/lib/python3.7/site-packages/torch/nn/modules/module.py", line 593, in _apply
    param_applied = fn(param)
  File "/home/ccbien/.conda/envs/point-mlp/lib/python3.7/site-packages/torch/nn/modules/module.py", line 897, in convert
    return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking)
RuntimeError: CUDA error: the provided PTX was compiled with an unsupported toolchain.
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.

The following is the output of command ‘nvidia-smi’:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.45.01    Driver Version: 455.45.01    CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce RTX 3090    On   | 00000000:01:00.0 Off |                  N/A |
| 45%   53C    P0   141W / 350W |      1MiB / 24268MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

I’ve tried with different versions of pytorch, cudatoolkit (10.2, 11.3) but got the same error.

[etaoxing]

I’ve also run into this issue. To fix this, I removed the dependency on pointnet2_ops, since it was only used once: https://github.com/ma-xu/pointMLP-pytorch/blob/c1d6235405a8e53027d5afa1349a368788fa2469/classification_ModelNet40/models/pointmlp.py#L159-L170

For any interested:

        from pytorch3d.ops import sample_farthest_points, knn_points
        # could also switch to pytorch_geometric

        # fps_idx = torch.multinomial(torch.linspace(0, N - 1, steps=N).repeat(B, 1).to(xyz.device), num_samples=self.groups, replacement=False).long()
        fps_idx = farthest_point_sample(xyz, self.groups).long()
        # fps_idx = pointnet2_utils.furthest_point_sample(xyz, self.groups).long()  # [B, npoint]
        new_xyz, fps_idx = sample_farthest_points(xyz, K=self.groups)
        # new_xyz = index_points(xyz, fps_idx)  # [B, npoint, 3]
        new_points = index_points(points, fps_idx)  # [B, npoint, d]

        # idx = knn_point(self.kneighbors, xyz, new_xyz)
        _, idx, _ = knn_points(new_xyz, xyz, K=self.kneighbors, return_nn=False)
        # idx = query_ball_point(radius, nsample, xyz, new_xyz)
        grouped_points = index_points(points, idx)  # [B, npoint, k, d]

        if self.use_xyz:
            grouped_xyz = index_points(xyz, idx)  # [B, npoint, k, 3]
            grouped_points = torch.cat([grouped_points, grouped_xyz],dim=-1)  # [B, npoint, k, d+3]

[ma-xu]

@etaoxing Thanks a lot!