error when running test script:

[Spartanzhao]

I had installed the graph gym step by step as readme mentioned, But I got this error:
Traceback (most recent call last):
File "main.py", line 5, in
from torch_geometric import seed_everything
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/init.py", line 1, in
import torch_geometric.utils
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/utils/init.py", line 3, in
from .scatter import scatter
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/utils/scatter.py", line 7, in
import torch_geometric.typing
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/typing.py", line 37, in
import torch_sparse # noqa
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/init.py", line 40, in
from .tensor import SparseTensor # noqa
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/tensor.py", line 13, in
class SparseTensor(object):
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py", line 974, in script
_compile_and_register_class(obj, _rcb, qualified_name)
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py", line 67, in _compile_and_register_class
torch._C._jit_script_class_compile(qualified_name, ast, defaults, rcb)
RuntimeError:
Tried to access nonexistent attribute or method 'crow_indices' of type 'Tensor'.:
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/tensor.py", line 109
def from_torch_sparse_csr_tensor(self, mat: torch.Tensor,
has_value: bool = True):
rowptr = mat.crow_indices()
~~~~~~~~~~~~~~~~ <--- HERE
col = mat.col_indices()

How Can I solve this issue?

[Spartanzhao]

Here is the answer from gpt4:
The error message indicates that the torch_sparse package is attempting to access a method crow_indices on a torch.Tensor object which does not exist. This could be due to several reasons:
Version Incompatibility: The version of PyTorch you're using might not be compatible with the version of torch_sparse. The crow_indices method might have been introduced in a newer version of PyTorch, and if you're using an older version, this method would not be available.
Incorrect Tensor Type: The function from_torch_sparse_csr_tensor is expecting a tensor of type CSR (Compressed Sparse Row), which should have the method crow_indices. If the tensor mat is not of the correct sparse format, this method would not be present.
To solve this issue, consider the following steps:
Update PyTorch: Ensure that you're using the latest version of PyTorch that supports CSR format and has the crow_indices method for sparse tensors.
bash
复制
pip install --upgrade torch
Check Tensor Format: Make sure that the tensor you're providing to from_torch_sparse_csr_tensor is indeed a CSR tensor. In PyTorch, CSR tensors are created with torch.sparse_csr_tensor.
Compatibility Check: Verify that all related libraries (torch_geometric, torch_sparse, and torch) are compatible with each other. Sometimes, libraries are updated to leverage new features in dependencies, and using mismatched versions can cause errors.
Environment Clean-Up: If you're working in a conda environment and encountering package conflicts, it may be helpful to create a new environment from scratch, ensuring that you install packages that are compatible with each other.
bash

conda create -n new_env_name python=3.7
conda activate new_env_name
pip install torch torch_geometric torch_sparse
Please replace new_env_name with your desired environment name and adjust the Python version if needed.
If after these steps you're still facing issues, it could be helpful to check the documentation of torch_sparse or open an issue on the relevant GitHub repository for assistance from the developers.

Is this correct answer?
I have been tried to install graphgym for month, and called some professional helpers, but still don't work, what other information I need provide to you, then you can help me to install the graphgym successfully?

[Spartanzhao]

When I am import these torch packages, I got this error:
In [1]: import torch_geometric

RuntimeError Traceback (most recent call last)
in
----> 1 import torch_geometric

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/init.py in
----> 1 import torch_geometric.utils
2 import torch_geometric.data
3 import torch_geometric.sampler
4 import torch_geometric.loader
5 import torch_geometric.transforms

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/utils/init.py in
1 import copy
2
----> 3 from .scatter import scatter
4 from .segment import segment
5 from .sort import index_sort

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/utils/scatter.py in
5 from torch import Tensor
6
----> 7 import torch_geometric.typing
8 from torch_geometric.typing import torch_scatter
9

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_geometric/typing.py in
35
36 try:
---> 37 import torch_sparse # noqa
38 from torch_sparse import SparseTensor
39 WITH_TORCH_SPARSE = True

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/init.py in
38
39 from .storage import SparseStorage # noqa
---> 40 from .tensor import SparseTensor # noqa
41 from .transpose import t # noqa
42 from .narrow import narrow, narrow_diag # noqa

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/tensor.py in
11
12 @torch.jit.script
---> 13 class SparseTensor(object):
14 storage: SparseStorage
15

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py in script(obj, optimize, _frames_up, _rcb)
972 if _rcb is None:
973 _rcb = _jit_internal.createResolutionCallbackFromFrame(_frames_up + 1)
--> 974 _compile_and_register_class(obj, _rcb, qualified_name)
975 return obj
976 else:

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py in _compile_and_register_class(obj, rcb, qualified_name)
65 ast = get_jit_class_def(obj, obj.name)
66 defaults = torch.jit.frontend.get_default_args_for_class(obj)
---> 67 torch._C._jit_script_class_compile(qualified_name, ast, defaults, rcb)
68 torch.jit._state._add_script_class(obj, qualified_name)
69

RuntimeError:
Tried to access nonexistent attribute or method 'crow_indices' of type 'Tensor'.:
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/tensor.py", line 109
def from_torch_sparse_csr_tensor(self, mat: torch.Tensor,
has_value: bool = True):
rowptr = mat.crow_indices()
~~~~~~~~~~~~~~~~ <--- HERE
col = mat.col_indices()

In [2]: import torch_scatter

In [3]: import torch_sparse

RuntimeError Traceback (most recent call last)
in
----> 1 import torch_sparse

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/init.py in
38
39 from .storage import SparseStorage # noqa
---> 40 from .tensor import SparseTensor # noqa
41 from .transpose import t # noqa
42 from .narrow import narrow, narrow_diag # noqa

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_sparse/tensor.py in
11
12 @torch.jit.script
---> 13 class SparseTensor(object):
14 storage: SparseStorage
15

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py in script(obj, optimize, _frames_up, _rcb)
972 if _rcb is None:
973 _rcb = _jit_internal.createResolutionCallbackFromFrame(_frames_up + 1)
--> 974 _compile_and_register_class(obj, _rcb, qualified_name)
975 return obj
976 else:

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py in _compile_and_register_class(obj, rcb, qualified_name)
65 ast = get_jit_class_def(obj, obj.name)
66 defaults = torch.jit.frontend.get_default_args_for_class(obj)
---> 67 torch._C._jit_script_class_compile(qualified_name, ast, defaults, rcb)
68 torch.jit._state._add_script_class(obj, qualified_name)
69

RuntimeError: class 'torch.torch_sparse.tensor.SparseTensor' already defined.

In [4]: import torch_cluster

RuntimeError Traceback (most recent call last)
in
----> 1 import torch_cluster

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_cluster/init.py in
43 from .nearest import nearest # noqa
44 from .radius import radius, radius_graph # noqa
---> 45 from .rw import random_walk # noqa
46 from .sampler import neighbor_sampler # noqa
47

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_cluster/rw.py in
16 num_nodes: Optional[int] = None,
17 return_edge_indices: bool = False,
---> 18 ) -> Union[Tensor, Tuple[Tensor, Tensor]]:
19 """Samples random walks of length :obj:walk_length from all node indices
20 in :obj:start in the graph given by :obj:(row, col) as described in the

~/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch/jit/_script.py in script(obj, optimize, _frames_up, _rcb)
988 _rcb = _jit_internal.createResolutionCallbackFromClosure(obj)
989 fn = torch._C._jit_script_compile(
--> 990 qualified_name, ast, _rcb, get_default_args(obj)
991 )
992 # Forward docstrings

RuntimeError:
General Union types are not currently supported. Only Union[T, NoneType] (i.e. Optional[T]) is supported.:
File "/data02/home/scv9476/.conda/envs/graphgym_env/lib/python3.7/site-packages/torch_cluster/rw.py", line 18
num_nodes: Optional[int] = None,
return_edge_indices: bool = False,
) -> Union[Tensor, Tuple[Tensor, Tensor]]:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
"""Samples random walks of length :obj:walk_length from all node indices
in :obj:start in the graph given by :obj:(row, col) as described in the
I think may be these torch packages are incompatible ? Do you know any way that I can install these packages which are not incompatible at all and can run your graphgym?