Fix full test failures in master

pyproject.toml

@@ -182,6 +182,7 @@ quote-style = "single"
 convention = "google"
 
 [tool.pytest.ini_options]
+testpaths = ["test"]
 addopts = [
     "--capture=no",
     "--color=yes",
@@ -207,6 +208,8 @@ filterwarnings = [
     "error::DeprecationWarning",
     # TODO(rishipuri98): Remove usage of `torch_geometric.distributed` from `torch_geometric.llm`
     "ignore:.*torch_geometric.distributed.*:DeprecationWarning",
+    # Ignore due to pytorch-lightning using a deprecated pkg_resources API:
+    "ignore:.*Deprecated call to `pkg_resources.declare_namespace.*:DeprecationWarning",
 ]
 markers = [
     "rag: mark test as RAG test",

test/explain/algorithm/test_attention_explainer.py

@@ -1,6 +1,9 @@
+import os
+
 import pytest
 import torch
 
+import torch_geometric.typing
 from torch_geometric.explain import (
     AttentionExplainer,
     Explainer,
@@ -180,6 +183,10 @@ def forward(self, x_dict, edge_index_dict):
 batch = torch.tensor([0, 0, 0, 1, 1, 2, 2, 2])
 
 
+@pytest.mark.skipif(
+    os.name == 'nt' and torch_geometric.typing.WITH_PT24,
+    reason="Unknown heap corruption issue on Windows with PyTorch 2.4",
+)
 @pytest.mark.parametrize('index', [None, 2, torch.arange(3)])
 def test_attention_explainer(index, check_explanation):
     explainer = Explainer(
@@ -216,6 +223,10 @@ def test_attention_explainer_supports(explanation_type, node_mask_type):
         )
 
 
+@pytest.mark.skipif(
+    os.name == 'nt' and torch_geometric.typing.WITH_PT24,
+    reason="Unknown heap corruption issue on Windows with PyTorch 2.4",
+)
 def test_attention_explainer_attentive_fp(check_explanation):
     model = AttentiveFP(3, 16, 1, edge_dim=5, num_layers=2, num_timesteps=2)
 
@@ -235,6 +246,10 @@ def test_attention_explainer_attentive_fp(check_explanation):
     check_explanation(explanation, None, explainer.edge_mask_type)
 
 
+@pytest.mark.skipif(
+    os.name == 'nt' and torch_geometric.typing.WITH_PT24,
+    reason="Unknown heap corruption issue on Windows with PyTorch 2.4",
+)
 @pytest.mark.parametrize('index', [None, 2, torch.arange(3)])
 def test_attention_explainer_hetero(index, hetero_data,
                                     check_explanation_hetero):

test/nn/conv/test_point_transformer_conv.py

@@ -3,11 +3,13 @@
 
 import torch_geometric.typing
 from torch_geometric.nn import PointTransformerConv
-from torch_geometric.testing import is_full_test
+from torch_geometric.testing import is_full_test, withPackage
 from torch_geometric.typing import SparseTensor
 from torch_geometric.utils import to_torch_csc_tensor
 
 
+# Skip on PyTorch 1.13 due to numerical instability:
+@withPackage('torch>=2.0.0')
 def test_point_transformer_conv():
     x1 = torch.rand(4, 16)
     x2 = torch.randn(2, 8)

test/nn/conv/test_sage_conv.py

@@ -126,7 +126,7 @@ def test_multi_aggr_sage_conv(aggr_kwargs):
 
 @withDevice
 @onlyLinux
-@withPackage('torch>=2.1.0')
+@withPackage('torch>=2.1.0', 'torch!=2.3.0', 'torch!=2.3.1')
 def test_compile_multi_aggr_sage_conv(device):
     import torch._dynamo as dynamo
 

test/nn/models/test_basic_gnn.py

@@ -204,7 +204,7 @@ def test_basic_gnn_inference(get_dataset, jk):
 @withDevice
 @onlyLinux
 @onlyFullTest
-@withPackage('torch>=2.0.0')
+@withPackage('torch>=2.0.0', 'torch!=2.3.0', 'torch!=2.3.1')
 def test_compile_basic(device):
     x = torch.randn(3, 8, device=device)
     edge_index = torch.tensor([[0, 1, 1, 2], [1, 0, 2, 1]], device=device)

test/nn/models/test_lpformer.py

@@ -5,7 +5,7 @@
 from torch_geometric.utils import to_undirected
 
 
-@withPackage('numba')  # For ppr calculation
+@withPackage('numba', 'torch>=2.0.0')  # For ppr calculation
 def test_lpformer():
     model = LPFormer(16, 32, num_gnn_layers=2, num_transformer_layers=1)
     assert str(

test/nn/test_compile_basic.py

@@ -45,7 +45,7 @@ def fused_gather_scatter(x, edge_index, reduce=('sum', 'mean', 'max')):
 @withDevice
 @onlyLinux
 @onlyFullTest
-@withPackage('torch>=2.0.0')
+@withPackage('torch>=2.0.0', 'torch!=2.3.0', 'torch!=2.3.1')
 def test_torch_compile(device):
     x = torch.randn(10, 16, device=device)
     edge_index = torch.randint(0, x.size(0), (2, 40), device=device)

test/nn/test_compile_conv.py

@@ -29,7 +29,7 @@ def forward(self, x: Tensor, edge_index: Tensor) -> Tensor:
 @withDevice
 @onlyLinux
 @onlyFullTest
-@withPackage('torch>=2.1.0')
+@withPackage('torch>=2.1.0', 'torch!=2.3.0', 'torch!=2.3.1')
 @pytest.mark.parametrize('Conv', [GCNConv, SAGEConv])
 def test_compile_conv(device, Conv):
     import torch._dynamo as dynamo
@@ -52,7 +52,7 @@ def test_compile_conv(device, Conv):
 @withDevice
 @onlyLinux
 @onlyFullTest
-@withPackage('torch==2.3')
+@withPackage('torch>=2.4.0')
 @pytest.mark.parametrize('Conv', [GCNConv, SAGEConv])
 def test_compile_conv_edge_index(device, Conv):
     import torch._dynamo as dynamo

test/nn/test_compile_dynamic.py

@@ -28,7 +28,7 @@ def forward(self, x: Tensor, edge_index: Tensor) -> Tensor:
 @withDevice
 @onlyLinux
 @onlyFullTest
-@withPackage('torch>2.0.0')
+@withPackage('torch>2.0.0', 'torch!=2.3.0', 'torch!=2.3.1')
 def test_dynamic_torch_compile(device):
     conv = MySAGEConv(64, 64).to(device)
     conv = torch.compile(conv, dynamic=True)

torch_geometric/data/extract.py

@@ -28,7 +28,10 @@ def extract_tar(
     """
     maybe_log(path, log)
     with tarfile.open(path, mode) as f:
-        f.extractall(folder, filter='data')
+        if sys.version_info >= (3, 12):
+            f.extractall(folder, filter='data')
+        else:
+            f.extractall(folder)
 
 
 def extract_zip(path: str, folder: str, log: bool = True) -> None: