Tensor Size Mismatch

[kevintsq]

The input image is opened from KITTI. The intrinsics is a 3x3 matrix. Why is there a tensor size mismatch?

img = Image.open(img_path).convert('RGB')
rgb = torch.tensor(np.asarray(img), device=self.device).permute(2, 0, 1).unsqueeze(0) / 255.
depth = self.depth_model(rgb, self.intrinsics.unsqueeze(0))

Traceback (most recent call last):
  File "depth_predictor.py", line 50, in predict_depth
    depth = self.depth_model(rgb, self.intrinsics.unsqueeze(0))
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "anaconda3/envs/cfgs/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".cache/torch/hub/TRI-ML_vidar_main/vidar/arch/networks/perceiver/ZeroDepthNet.py", line 765, in forward
    encoded_data = self.encode(
                   ^^^^^^^^^^^^
  File ".cache/torch/hub/TRI-ML_vidar_main/vidar/arch/networks/perceiver/ZeroDepthNet.py", line 465, in encode
    embeddings = [self.encode_embeddings(data, embeddings, scene, idx=i)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".cache/torch/hub/TRI-ML_vidar_main/vidar/arch/networks/perceiver/ZeroDepthNet.py", line 465, in <listcomp>
    embeddings = [self.encode_embeddings(data, embeddings, scene, idx=i)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".cache/torch/hub/TRI-ML_vidar_main/vidar/arch/networks/perceiver/ZeroDepthNet.py", line 498, in encode_embeddings
    embeddings = self.merge_embeddings(embeddings, sources)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".cache/torch/hub/TRI-ML_vidar_main/vidar/arch/networks/perceiver/ZeroDepthNet.py", line 305, in merge_embeddings
    cat = torch.cat(cat, -1)
          ^^^^^^^^^^^^^^^^^^
RuntimeError: Sizes of tensors must match except in dimension 2. Expected size 28830 but got size 29234 for tensor number 1 in the list.