AttributeError: type object 'DCNv3' has no attribute 'dcnv3_forward'

[MayoMathur]

Hi , I was trying to create a separate object of DCNv3

dcn = DCNv3(
            channels=3,
            kernel_size=3,
            stride=1,
            pad=1,
            dilation=1,
            group=1,)
x = torch.randn(2, 64, 64, 3)

but I encountered the following error

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
[/home/thor/pythonProjects/InternImage/segmentation/test_notebook.ipynb](https://file/+.[vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/test_notebook.ipynb](http://vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/test_notebook.ipynb)) Cell 9 line 1
----> [1](vscode-notebook-cell:/home/thor/pythonProjects/InternImage/segmentation/test_notebook.ipynb#X11sZmlsZQ%3D%3D?line=0) dcn(x.permute(0,2,3,1))

File [~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/nn/modules/module.py:1501](https://file/+.[vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/nn/modules/module.py:1501](http://vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/nn/modules/module.py:1501)), in Module._call_impl(self, *args, **kwargs)
   1496 # If we don't have any hooks, we want to skip the rest of the logic in
   1497 # this function, and just call forward.
   1498 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1499         or _global_backward_pre_hooks or _global_backward_hooks
   1500         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501     return forward_call(*args, **kwargs)
   1502 # Do not call functions when jit is used
   1503 full_backward_hooks, non_full_backward_hooks = [], []

File [~/pythonProjects/InternImage/segmentation/ops_dcnv3/modules/dcnv3_base.py:272](https://file/+.[vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/pythonProjects/InternImage/segmentation/ops_dcnv3/modules/dcnv3_base.py:272](http://vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/pythonProjects/InternImage/segmentation/ops_dcnv3/modules/dcnv3_base.py:272)), in DCNv3.forward(self, input)
    269 mask = self.mask(x1).reshape(N, H, W, self.group, -1)
    270 mask = F.softmax(mask, -1).reshape(N, H, W, -1).type(dtype)
--> 272 x = DCNv3Function.apply(
    273     x, offset, mask,
    274     self.kernel_size, self.kernel_size,
    275     self.stride, self.stride,
    276     self.pad, self.pad,
    277     self.dilation, self.dilation,
    278     self.group, self.group_channels,
    279     self.offset_scale,
    280     256)
    282 if self.center_feature_scale:
    283     center_feature_scale = self.center_feature_scale_module(
    284         x1, self.center_feature_scale_proj_weight, self.center_feature_scale_proj_bias)

File [~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/autograd/function.py:506](https://file/+.[vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/autograd/function.py:506](http://vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/autograd/function.py:506)), in Function.apply(cls, *args, **kwargs)
    503 if not torch._C._are_functorch_transforms_active():
    504     # See NOTE: [functorch vjp and autograd interaction]
    505     args = _functorch.utils.unwrap_dead_wrappers(args)
--> 506     return super().apply(*args, **kwargs)  # type: ignore[misc]
    508 if cls.setup_context == _SingleLevelFunction.setup_context:
    509     raise RuntimeError(
    510         'In order to use an autograd.Function with functorch transforms '
    511         '(vmap, grad, jvp, jacrev, ...), it must override the setup_context '
    512         'staticmethod. For more details, please see '
    513         'https://pytorch.org/docs/master/notes/extending.func.html')

File [~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/cuda/amp/autocast_mode.py:98](https://file/+.[vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/cuda/amp/autocast_mode.py:98](http://vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/anaconda3/envs/torchenv/lib/python3.9/site-packages/torch/cuda/amp/autocast_mode.py:98)), in custom_fwd.<locals>.decorate_fwd(*args, **kwargs)
     96 if cast_inputs is None:
     97     args[0]._fwd_used_autocast = torch.is_autocast_enabled()
---> 98     return fwd(*args, **kwargs)
     99 else:
    100     autocast_context = torch.is_autocast_enabled()

File [~/pythonProjects/InternImage/segmentation/ops_dcnv3/modules/dcnv3_base.py:39](https://file/+.[vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/pythonProjects/InternImage/segmentation/ops_dcnv3/modules/dcnv3_base.py:39](http://vscode-resource.vscode-cdn.net/home/thor/pythonProjects/InternImage/segmentation/~/pythonProjects/InternImage/segmentation/ops_dcnv3/modules/dcnv3_base.py:39)), in DCNv3Function.forward(ctx, input, offset, mask, kernel_h, kernel_w, stride_h, stride_w, pad_h, pad_w, dilation_h, dilation_w, group, group_channels, offset_scale, im2col_step)
     37 ctx.offset_scale = offset_scale
     38 ctx.im2col_step = im2col_step
---> 39 output = DCNv3.dcnv3_forward(
     40     input, offset, mask, kernel_h,
     41     kernel_w, stride_h, stride_w, pad_h,
     42     pad_w, dilation_h, dilation_w, group,
     43     group_channels, offset_scale, ctx.im2col_step)
     44 ctx.save_for_backward(input, offset, mask)
     46 return output

AttributeError: type object 'DCNv3' has no attribute 'dcnv3_forward'




Kindly guide me what are the steps that are needed to be taken. 

[leij0318]

You should have not compiled DCNv3! The dcnv3_forward and dcnv3_backward functions are written in C++ and finally compiled into DCNv3.pyd.

So you should use the command below to compile.

python setup.py build_ext install

After loading the DLL when the code is running, you can use DCNv3. Of course, if you do not compile, you can only use the dcnv3_pytorch function, and the inference speed may be slower.