Fix broken datasets url

CHANGELOG.md

@@ -15,18 +15,12 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ### Fixed
 
+- Fixed broken datasets url ([#10415](https://github.com/pyg-team/pytorch_geometric/pull/10415))
+
 ### Security
 
 ## [2.7.0] - 2025-10-14
 
-### Fixed
-
-- Fixed `ogbn_train_cugraph` example for distributed cuGraph ([#10439](https://github.com/pyg-team/pytorch_geometric/pull/10439))
-- Added `safe_onnx_export` function with workarounds for `onnx_ir.serde.SerdeError` issues in ONNX export ([#10422](https://github.com/pyg-team/pytorch_geometric/pull/10422))
-- Fixed importing PyTorch Lightning in `torch_geometric.graphgym` and `torch_geometric.data.lightning` when using `lightning` instead of `pytorch-lightning` ([#10404](https://github.com/pyg-team/pytorch_geometric/pull/10404), [#10417](https://github.com/pyg-team/pytorch_geometric/pull/10417)))
-- Fixed `detach()` warnings in example scripts involving tensor conversions ([#10357](https://github.com/pyg-team/pytorch_geometric/pull/10357))
-- Fixed non-tuple indexing to resolve PyTorch deprecation warning ([#10389](https://github.com/pyg-team/pytorch_geometric/pull/10389))
-
 ### Added
 
 - Added llm generated explanations to `TAGDataset` ([#9918](https://github.com/pyg-team/pytorch_geometric/pull/9918))
@@ -106,6 +100,11 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ### Fixed
 
+- Fixed `ogbn_train_cugraph` example for distributed cuGraph ([#10439](https://github.com/pyg-team/pytorch_geometric/pull/10439))
+- Fixed importing PyTorch Lightning in `torch_geometric.graphgym` and `torch_geometric.data.lightning` when using `lightning` instead of `pytorch-lightning` ([#10404](https://github.com/pyg-team/pytorch_geometric/pull/10404), [#10417](https://github.com/pyg-team/pytorch_geometric/pull/10417)))
+- Added `safe_onnx_export` function with workarounds for `onnx_ir.serde.SerdeError` issues in ONNX export ([#10422](https://github.com/pyg-team/pytorch_geometric/pull/10422))
+- Fixed `detach()` warnings in example scripts involving tensor conversions ([#10357](https://github.com/pyg-team/pytorch_geometric/pull/10357))
+- Fixed non-tuple indexing to resolve PyTorch deprecation warning ([#10389](https://github.com/pyg-team/pytorch_geometric/pull/10389))
 - Fixed conversion to/from `cuGraph` graph objects by ensuring `cudf` column names are correctly specified ([#10343](https://github.com/pyg-team/pytorch_geometric/pull/10343))
 - Fixed `_recursive_config()` for `torch.nn.ModuleList` and `torch.nn.ModuleDict` ([#10124](https://github.com/pyg-team/pytorch_geometric/pull/10124), [#10129](https://github.com/pyg-team/pytorch_geometric/pull/10129))
 - Fixed the `k_hop_subgraph()` method for directed graphs ([#9756](https://github.com/pyg-team/pytorch_geometric/pull/9756))

test/datasets/test_deezer_europe.py

@@ -0,0 +1,11 @@
+from torch_geometric.datasets import DeezerEurope
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+def test_deezer_europe():
+    dataset = DeezerEurope(root='./data/DeezerEurope')
+    data = dataset[0]
+    assert data.x.size() == (28281, 128)
+    assert data.edge_index.size() == (2, 92752)

test/datasets/test_facebook.py

@@ -0,0 +1,11 @@
+from torch_geometric.datasets import FacebookPagePage
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+def test_facebook():
+    dataset = FacebookPagePage(root='./data/FacebookPagePage')
+    data = dataset[0]
+    assert data.x.size() == (22470, 128)
+    assert data.edge_index.size() == (2, 171002)

test/datasets/test_gemsec.py

@@ -0,0 +1,10 @@
+from torch_geometric.datasets import GemsecDeezer
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+def test_gemsec():
+    dataset = GemsecDeezer(root='./data/GemsecDeezer', name='HU')
+    data = dataset[0]
+    assert data.edge_index.size() == (2, 222887)

test/datasets/test_github.py

@@ -0,0 +1,11 @@
+from torch_geometric.datasets import GitHub
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+def test_github():
+    dataset = GitHub(root='./data/GitHub')
+    data = dataset[0]
+    assert data.x.size() == (37700, 128)
+    assert data.edge_index.size() == (2, 289003)

test/datasets/test_lastfm_asia.py

@@ -0,0 +1,11 @@
+from torch_geometric.datasets import LastFMAsia
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+def test_lastfm_asia():
+    dataset = LastFMAsia(root='./data/LastFMAsia')
+    data = dataset[0]
+    assert data.x.size() == (7624, 128)
+    assert data.edge_index.size() == (2, 27806)

test/datasets/test_twitch.py

@@ -0,0 +1,21 @@
+import pytest
+
+from torch_geometric.datasets import Twitch
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+@pytest.mark.parametrize('name, n_dim, e_dim', [
+    ('ENGB', 7126, 35324),
+    ('ES', 4648, 59382),
+    ('FR', 6551, 112666),
+    ('PTBR', 1912, 31299),
+    ('RU', 4385, 37304),
+    ('DE', 9498, 153138),
+])
+def test_twitch(name: str, n_dim: int, e_dim: int):
+    dataset = Twitch(root='./data/Twitch', name=name)
+    data = dataset[0]
+    assert data.x.size() == (n_dim, 128)
+    assert data.edge_index.size() == (2, e_dim)

test/datasets/test_wikipedia_network.py

@@ -0,0 +1,12 @@
+from torch_geometric.datasets import WikipediaNetwork
+from torch_geometric.testing import onlyFullTest, onlyOnline
+
+
+@onlyOnline
+@onlyFullTest
+def test_wikipedia_network():
+    dataset = WikipediaNetwork(root='./data/WikipediaNetwork',
+                               name='crocodile')
+    data = dataset[0]
+    assert data.x.size() == (11631, 128)
+    assert data.edge_index.size() == (2, 180020)

torch_geometric/datasets/deezer_europe.py

@@ -1,9 +1,14 @@
-from typing import Callable, Optional
+from typing import Callable, List, Optional
 
 import numpy as np
 import torch
 
-from torch_geometric.data import Data, InMemoryDataset, download_url
+from torch_geometric.data import (
+    Data,
+    InMemoryDataset,
+    download_url,
+    extract_zip,
+)
 
 
 class DeezerEurope(InMemoryDataset):
@@ -28,7 +33,7 @@ class DeezerEurope(InMemoryDataset):
             (default: :obj:`False`)
     """
 
-    url = 'https://graphmining.ai/datasets/ptg/deezer_europe.npz'
+    url = 'https://snap.stanford.edu/data/deezer_europe.zip'
 
     def __init__(
         self,
@@ -42,21 +47,44 @@ def __init__(
         self.load(self.processed_paths[0])
 
     @property
-    def raw_file_names(self) -> str:
-        return 'deezer_europe.npz'
+    def raw_file_names(self) -> List[str]:
+        return [
+            f'deezer_europe/{x}' for x in [
+                'deezer_europe_edges.csv',
+                'deezer_europe_features.json',
+                'deezer_europe_target.csv',
+            ]
+        ]
 
     @property
     def processed_file_names(self) -> str:
         return 'data.pt'
 
     def download(self) -> None:
-        download_url(self.url, self.raw_dir)
+        file_path = download_url(self.url, self.raw_dir)
+        extract_zip(file_path, self.raw_dir)
 
     def process(self) -> None:
-        data = np.load(self.raw_paths[0], 'r', allow_pickle=True)
-        x = torch.from_numpy(data['features']).to(torch.float)
-        y = torch.from_numpy(data['target']).to(torch.long)
-        edge_index = torch.from_numpy(data['edges']).to(torch.long)
+        import json
+
+        import pandas as pd
+        edges = pd.read_csv(self.raw_paths[0], dtype=int)
+        features = json.load(open(self.raw_paths[1]))
+        target = pd.read_csv(self.raw_paths[2], dtype=int)
+
+        xs = []
+        n_feats = 128
+        for i in target['id'].values:
+            f = [0] * n_feats
+            if str(i) in features:
+                n_len = len(features[str(i)])
+                f = features[str(
+                    i)][:n_feats] if n_len >= n_feats else features[str(
+                        i)] + [0] * (n_feats - n_len)
+            xs.append(f)
+        x = torch.from_numpy(np.array(xs)).to(torch.float)
+        y = torch.from_numpy(target.values[:, 1]).t().to(torch.long)
+        edge_index = torch.from_numpy(edges.values).to(torch.long)
         edge_index = edge_index.t().contiguous()
 
         data = Data(x=x, y=y, edge_index=edge_index)

torch_geometric/datasets/facebook.py

@@ -1,9 +1,14 @@
-from typing import Callable, Optional
+from typing import Callable, List, Optional
 
 import numpy as np
 import torch
 
-from torch_geometric.data import Data, InMemoryDataset, download_url
+from torch_geometric.data import (
+    Data,
+    InMemoryDataset,
+    download_url,
+    extract_zip,
+)
 
 
 class FacebookPagePage(InMemoryDataset):
@@ -27,7 +32,7 @@ class FacebookPagePage(InMemoryDataset):
             (default: :obj:`False`)
     """
 
-    url = 'https://graphmining.ai/datasets/ptg/facebook.npz'
+    url = 'https://snap.stanford.edu/data/facebook_large.zip'
 
     def __init__(
         self,
@@ -41,21 +46,47 @@ def __init__(
         self.load(self.processed_paths[0])
 
     @property
-    def raw_file_names(self) -> str:
-        return 'facebook.npz'
+    def raw_file_names(self) -> List[str]:
+        return [
+            f'facebook_large/{x}' for x in [
+                'musae_facebook_edges.csv',
+                'musae_facebook_features.json',
+                'musae_facebook_target.csv',
+            ]
+        ]
 
     @property
     def processed_file_names(self) -> str:
         return 'data.pt'
 
     def download(self) -> None:
-        download_url(self.url, self.raw_dir)
+        file_path = download_url(self.url, self.raw_dir)
+        extract_zip(file_path, self.raw_dir)
 
     def process(self) -> None:
-        data = np.load(self.raw_paths[0], 'r', allow_pickle=True)
-        x = torch.from_numpy(data['features']).to(torch.float)
-        y = torch.from_numpy(data['target']).to(torch.long)
-        edge_index = torch.from_numpy(data['edges']).to(torch.long)
+        import json
+
+        import pandas as pd
+        edges = pd.read_csv(self.raw_paths[0], dtype=int)
+        features = json.load(open(self.raw_paths[1]))
+        target = pd.read_csv(self.raw_paths[2])
+        page_type = sorted(target['page_type'].value_counts().index.tolist())
+        target['y'] = target['page_type'].apply(lambda x: page_type.index(x))
+
+        # import pdb; pdb.set_trace()  # noqa: T201
+        xs = []
+        n_feats = 128
+        for i in target['id'].values:
+            f = [0] * n_feats
+            if str(i) in features:
+                n_len = len(features[str(i)])
+                f = features[str(
+                    i)][:n_feats] if n_len >= n_feats else features[str(
+                        i)] + [0] * (n_feats - n_len)
+            xs.append(f)
+        x = torch.from_numpy(np.array(xs)).to(torch.float)
+        y = torch.from_numpy(target['y'].values).t().to(torch.long)
+        edge_index = torch.from_numpy(edges.values).to(torch.long)
         edge_index = edge_index.t().contiguous()
 
         data = Data(x=x, y=y, edge_index=edge_index)

torch_geometric/datasets/gemsec.py

@@ -1,10 +1,14 @@
 import os.path as osp
-from typing import Callable, Optional
+from typing import Callable, List, Optional
 
-import numpy as np
 import torch
 
-from torch_geometric.data import Data, InMemoryDataset, download_url
+from torch_geometric.data import (
+    Data,
+    InMemoryDataset,
+    download_url,
+    extract_tar,
+)
 
 
 class GemsecDeezer(InMemoryDataset):
@@ -31,7 +35,7 @@ class GemsecDeezer(InMemoryDataset):
             (default: :obj:`False`)
     """
 
-    url = 'https://graphmining.ai/datasets/ptg/gemsec'
+    url = 'https://snap.stanford.edu/data/gemsec_deezer_dataset.tar.gz'
 
     def __init__(
         self,
@@ -56,23 +60,31 @@ def processed_dir(self) -> str:
         return osp.join(self.root, self.name, 'processed')
 
     @property
-    def raw_file_names(self) -> str:
-        return f'{self.name}.npz'
+    def raw_file_names(self) -> List[str]:
+        return [
+            f'deezer_clean_data/{x}' for x in [
+                f'{self.name}_edges.csv',
+                f'{self.name}_genres.json',
+            ]
+        ]
 
     @property
     def processed_file_names(self) -> str:
         return 'data.pt'
 
     def download(self) -> None:
-        download_url(osp.join(self.url, self.name + '.npz'), self.raw_dir)
+        file_path = download_url(self.url, self.raw_dir)
+        extract_tar(file_path, self.raw_dir)
 
     def process(self) -> None:
-        data = np.load(self.raw_paths[0], 'r', allow_pickle=True)
-        y = torch.from_numpy(data['target']).to(torch.long)
-        edge_index = torch.from_numpy(data['edges']).to(torch.long)
+
+        import pandas as pd
+        edges = pd.read_csv(self.raw_paths[0], dtype=int)
+
+        edge_index = torch.from_numpy(edges.values).to(torch.long)
         edge_index = edge_index.t().contiguous()
 
-        data = Data(y=y, edge_index=edge_index)
+        data = Data(edge_index=edge_index)
 
         if self.pre_transform is not None:
             data = self.pre_transform(data)