Cannot install with PyTorch Nightly

[sunshineatnoon]

📚 Installation

Environment

• OS: Ubuntu 16.04
• Python version: 3.7.9
• PyTorch version: 1.9.0.dev20210220
• CUDA/cuDNN version: 10.2.89
• GCC version: 5.4.0
• How did you try to install PyTorch Geometric and its extensions (wheel, source):

  conda install pytorch torchvision torchaudio -c pytorch-nightly
  pip install torch-scatter
  pip install torch-sparse
  pip install torch-cluster
  pip install torch-spline-conv
  pip install torch-geometric
  

• Any other relevant information:
  The pip installation is successful, but when I try import torch_sparse, I got the following error:

  Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/anaconda2/envs/svg/lib/python3.7/site-packages/torch_sparse/__init__.py", line 34, in <module>
    from .storage import SparseStorage  # noqa
  File "/home/anaconda2/envs/svg/lib/python3.7/site-packages/torch_sparse/storage.py", line 21, in <module>
    class SparseStorage(object):
  File "/home/xtli/anaconda2/envs/svg/lib/python3.7/site-packages/torch/jit/_script.py", line 979, in script
    _compile_and_register_class(obj, _rcb, qualified_name)
  File "/home/anaconda2/envs/svg/lib/python3.7/site-packages/torch/jit/_script.py", line 72, in _compile_and_register_class
    torch._C._jit_script_class_compile(qualified_name, ast, defaults, rcb)
  RuntimeError:
  attribute assignment is not defined on __torch__.torch_sparse.storage.SparseStorage:
    File "/home/anaconda2/envs/svg/lib/python3.7/site-packages/torch_sparse/storage.py", line 145
    def empty(self):
        self = SparseStorage.__new__(SparseStorage)
        self._row = None
        ~~~~~~~~~~~~~~~~ <--- HERE
        self._rowptr = None
        self._value = None
  

Checklist

• I followed the installation guide.
• I cannot find my error message in the FAQ.
• I set up CUDA correctly and can compile CUDA code via nvcc.
• I do have multiple CUDA versions on my machine.

[rusty1s]

This looks like some TorchScript internals changed in PyTorch master. We will fix that once a new official PyTorch release lands. We cannot fix that for now to guarantee correct behavior with the current PyTorch release.

[sunshineatnoon]

Got it. Thanks for the quick response and looking forward to the new release.