Source: pytorch-geometric
Version: 2.4.0-2
Severity: important
Control: affects -1 src:scipy
pytorch-geometric is failing tests with scipy 1.13 from experimental,
evidently due to use of a deprecated API.
The error message is
187s test/distributed/test_rpc.py On WorkerInfo(id=0, name=dist-feature-test-0):
187s AttributeError("'NoneType' object has no attribute 'rpc_async'")
187s Traceback (most recent call last):
187s File "/usr/lib/python3/dist-packages/torch/distributed/rpc/internal.py",
line 207, in _run_function
187s result = python_udf.func(*python_udf.args, **python_udf.kwargs)
187s ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
187s File
"/usr/lib/python3/dist-packages/torch_geometric/distributed/rpc.py", line 168,
in _rpc_async_call
187s return _rpc_call_pool.get(call_id).rpc_async(*args, **kwargs)
187s ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
187s AttributeError: 'NoneType' object has no attribute 'rpc_async'
187s
188s Process SpawnProcess-6:
188s Traceback (most recent call last):
188s File "/usr/lib/python3.11/multiprocessing/process.py", line 314, in
_bootstrap
188s self.run()
188s File "/usr/lib/python3.11/multiprocessing/process.py", line 108, in run
188s self._target(*self._args, **self._kwargs)
188s File
"/tmp/autopkgtest-lxc.glzbfzld/downtmp/autopkgtest_tmp/build/test/distributed/test_rpc.py",
line 79, in run_rpc_feature_test
188s assert torch.allclose(cpu_tensor0, tensor0.wait())
188s ^^^^^^^^^^^^^^
188s File "/usr/lib/python3/dist-packages/torch/futures/__init__.py", line
76, in wait
188s return super().wait()
188s ^^^^^^^^^^^^^^
188s File "/usr/lib/python3/dist-packages/torch/futures/__init__.py", line
268, in raise_error
188s raise fut_result
188s File
"/usr/lib/python3/dist-packages/torch_geometric/distributed/local_feature_store.py",
line 164, in when_finish
188s remote_feature_list = remote_fut.wait()
188s ^^^^^^^^^^^^^^^^^
188s File "/usr/lib/python3/dist-packages/torch/futures/__init__.py", line
76, in wait
188s return super().wait()
188s ^^^^^^^^^^^^^^
188s File "/usr/lib/python3/dist-packages/torch/futures/__init__.py", line
268, in raise_error
188s raise fut_result
188s File
"/usr/lib/python3/dist-packages/torch_geometric/distributed/local_feature_store.py",
line 256, in when_finish
188s fut_list = collect_fut.wait()
188s ^^^^^^^^^^^^^^^^^^
188s RuntimeError: AttributeError: On WorkerInfo(id=0,
name=dist-feature-test-0):
188s AttributeError("'NoneType' object has no attribute 'rpc_async'")
188s Traceback (most recent call last):
188s File "/usr/lib/python3/dist-packages/torch/distributed/rpc/internal.py",
line 207, in _run_function
188s result = python_udf.func(*python_udf.args, **python_udf.kwargs)
188s ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
188s File
"/usr/lib/python3/dist-packages/torch_geometric/distributed/rpc.py", line 168,
in _rpc_async_call
188s return _rpc_call_pool.get(call_id).rpc_async(*args, **kwargs)
188s ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
188s AttributeError: 'NoneType' object has no attribute 'rpc_async'
188s
188s
188s At:
188s /usr/lib/python3/dist-packages/torch/distributed/rpc/internal.py(234):
_handle_exception
Full log at
https://ci.debian.net/packages/p/pytorch-geometric/unstable/amd64/47741275/
scipy 1.12 has only recently been uploaded to debian, but upstream is
already preparing 1.14, so we can expect to upload 1.13 in the near
future.
