update v0.3.1

README.md

@@ -11,7 +11,7 @@
 % [![Documentation Status](http://mn.cs.tsinghua.edu.cn/autogl/documentation/?badge=latest)](http://mn.cs.tsinghua.edu.cn/autogl/documentation/?badge=latest)-->
 
 ## 最新消息
-
+- 2021.4.19 v0.3.1版本更新！首次更新中文教程！
 - 2021.12.31 v0.3.0-pre版本更新!
     - 智图目前支持[__Deep Graph Library (DGL)__](https://www.dgl.ai/)作为后端，以方便DGL的用户使用。目前在DGL后端已经支持同构图的节点分类、链接预测以及图分类等任务。智图现在也可兼容PyG 2.0版本。
     - 智图可以支持__异构图__节点分类任务!详情请参考[异构图教程](http://mn.cs.tsinghua.edu.cn/autogl/documentation/docfile/tutorial/t_hetero_node_clf.html)。

README_en.md

@@ -4,7 +4,7 @@ An autoML framework & toolkit for machine learning on graphs.
 
 *Actively under development by @THUMNLab*
 
-Feel free to open <a href="https://github.com/THUMNLab/AutoGL/issues">issues</a> or contact us at <a href="mailto:autogl@tsinghua.edu.cn">autogl@tsinghua.edu.cn</a> if you have any comments or suggestions!
+Feel free to open <a href="https://www.gitlink.org.cn/THUMNLab/AutoGL/issues">issues</a> or contact us at <a href="mailto:autogl@tsinghua.edu.cn">autogl@tsinghua.edu.cn</a> if you have any comments or suggestions!
 
 <!--
  [![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)
@@ -113,7 +113,7 @@ pip install autogl
 Run the following command to install this package from the source.
 
 ```
-git clone https://github.com/THUMNLab/AutoGL.git
+git clone https://gitlink.org.cn/THUMNLab/AutoGL.git
 cd AutoGL
 python setup.py install
 ```

autogl/__init__.py

@@ -16,4 +16,4 @@
     train,
 )
 
-__version__ = "0.3.0-pre"
+__version__ = "0.3.1"

autogl/module/feature/__init__.py

@@ -33,6 +33,9 @@
 from ._selectors import (
     FilterConstant, GBDTFeatureSelector
 )
+from ._auto_feature import (
+    IdentityFeature, OnlyConstFeature, AutoFeatureEngineer
+)
 
 __all__ = [
     "BaseFeatureEngineer",
@@ -61,5 +64,8 @@
     "NXGlobalEfficiency",
     "NXIsEulerian",
     "FilterConstant",
-    "GBDTFeatureSelector"
+    "GBDTFeatureSelector",
+    "IdentityFeature",
+    "OnlyConstFeature",
+    "AutoFeatureEngineer"
 ]

autogl/module/feature/_auto_feature.py

@@ -166,7 +166,7 @@ def __init__(
             verbosity: int = 0,
             *args, **kwargs
     ):
-        super(AutoFeatureEngineer, self).__init__(multi_graph=False)
+        super(AutoFeatureEngineer, self).__init__()
         self._ops = [op_sum, op_mean, op_max, op_min]
         self._sim = cosine_similarity
         self._fixlen = fix_length
@@ -208,7 +208,7 @@ def _fit(self, homogeneous_static_graph: autogl.data.graph.GeneralStaticGraph):
         for u, v in homogeneous_static_graph.edges.connections.t().numpy():
             neighbours[u].append(v)
         self.__neighbours: _typing.Sequence[np.ndarray] = tuple(
-            [np.ndarray(v) for v in neighbours]
+            [np.array(v) for v in neighbours]
         )
 
         x: np.ndarray = _original_features.numpy()

autogl/module/feature/_base_feature_engineer/_base_feature_engineer.py

@@ -29,7 +29,11 @@ def transform(self, dataset, inplace: bool = True):
         return dataset
 
     def fit_transform(self, dataset, inplace: bool = True):
-        return self.fit_transform(dataset, inplace)
+        for fe in self.fe_components:
+            dataset = fe.fit(dataset)
+        for fe in self.fe_components:
+            dataset = fe.transform(dataset)
+        return dataset
 
     def __init__(self, feature_engineers: _typing.Iterable[_AbstractBaseFeatureEngineer]):
         self.__fe_components: _typing.List[_AbstractBaseFeatureEngineer] = []

autogl/module/feature/_base_feature_engineer/_base_feature_engineer_dgl.py

@@ -40,7 +40,7 @@ def fit(self, dataset):
         with torch.no_grad():
             for i, data in enumerate(dataset):
                 dataset[i] = self.__postprocess(
-                    self._postprocess(self._transform(self._fit(self._preprocess(self.__preprocess(data)))))
+                    self._postprocess(self._fit(self._preprocess(self.__preprocess(data))))
                 )
         return dataset
 

autogl/module/feature/_base_feature_engineer/_base_feature_engineer_pyg.py

@@ -25,7 +25,7 @@ def fit(self, dataset):
         with torch.no_grad():
             for i, data in enumerate(dataset):
                 dataset[i] = self.__postprocess(
-                    self._postprocess(self._transform(self._fit(self._preprocess(self.__preprocess(data)))))
+                    self._postprocess(self._fit(self._preprocess(self.__preprocess(data))))
                 )
         return dataset
 

autogl/module/feature/_selectors/_basic.py

@@ -16,16 +16,12 @@ def __transform_homogeneous_static_graph(
     ) -> GeneralStaticGraph:
         if (
                 'x' in static_graph.nodes.data and
-                self._selection not in (Ellipsis, None) and
-                isinstance(self._selection, torch.Tensor) and
-                torch.is_tensor(self._selection) and self._selection.dim() == 1
+                isinstance(self._selection, (torch.Tensor, np.ndarray))
         ):
             static_graph.nodes.data['x'] = static_graph.nodes.data['x'][:, self._selection]
         if (
                 'feat' in static_graph.nodes.data and
-                self._selection not in (Ellipsis, None) and
-                isinstance(self._selection, torch.Tensor) and
-                torch.is_tensor(self._selection) and self._selection.dim() == 1
+                isinstance(self._selection, (torch.Tensor, np.ndarray))
         ):
             static_graph.nodes.data['feat'] = static_graph.nodes.data['feat'][:, self._selection]
         return static_graph

autogl/module/feature/_selectors/_gbdt.py

@@ -41,6 +41,7 @@ def _gbdt_generator(
     )
 
     num_classes: int = torch.max(data.y).item() + 1
+    parameters["num_class"] = num_classes
     __optimizer_parameters = {
         "num_boost_round": 100,
         "early_stopping_rounds": 5,
@@ -78,7 +79,7 @@ def _gbdt_generator(
         train_x = pd.DataFrame(x, columns=feature_index, index=None)
         dtrain = lightgbm.Dataset(train_x, label=label)
         clf = lightgbm.train(
-            train_set=dtrain, params=params,
+            train_set=dtrain, params=parameters,
             **__optimizer_parameters
         )
 

autogl/module/nas/algorithm/enas.py

@@ -82,7 +82,6 @@ def __init__(
         device="auto",
     ):
         super().__init__(device)
-        self.device = device
         self.num_epochs = num_epochs
         self.log_frequency = log_frequency
         self.entropy_weight = entropy_weight

autogl/module/nas/algorithm/gasso.py

@@ -61,7 +61,6 @@ def __init__(
         device="auto",
     ):
         super().__init__(device=device)
-        self.device = device
         self.num_epochs = num_epochs
         self.warmup_epochs = warmup_epochs
         self.model_lr = model_lr

autogl/module/nas/algorithm/rl.py

@@ -265,7 +265,6 @@ def __init__(
         disable_progress=False,
     ):
         super().__init__(device)
-        self.device = device
         self.num_epochs = num_epochs
         self.log_frequency = log_frequency
         self.entropy_weight = entropy_weight
@@ -447,7 +446,6 @@ def __init__(
         hardware_metric_limit=None,
     ):
         super().__init__(device)
-        self.device = device
         self.num_epochs = num_epochs
         self.log_frequency = log_frequency
         self.entropy_weight = entropy_weight

autogl/module/nas/algorithm/spos.py

@@ -228,7 +228,6 @@ def __init__(
         device="cuda",
     ):
         super().__init__(device)
-        self.device = device
         self.model_lr=5e-3
         self.model_wd=5e-4
         self.n_warmup = n_warmup

autogl/module/nas/space/autoattend.py

@@ -57,6 +57,9 @@ def __init__(
 
     ):
         super().__init__()
+
+        from autogl.backend import DependentBackend;assert not DependentBackend.is_dgl(),"Now AutoAttend only support pyg"
+
         self.layer_number = layer_number
         self.hidden_dim = hidden_dim
         self.input_dim = input_dim

autogl/solver/classifier/node_classifier.py

@@ -313,6 +313,7 @@ def fit(
         # check whether the dataset has features.
         # currently we only support graph classification with features.
 
+        graph_data = get_graph_from_dataset(dataset, 0)
         feat = get_graph_node_features(graph_data)
         assert feat is not None, (
             "Does not support fit on non node-feature dataset!"

docs/docfile/tutorial/t_backend.rst

@@ -9,13 +9,13 @@ enable users from both end benifiting the automation of graph learning.
 To specify one specific backend, you can declare the backend using environment variables
 ``AUTOGL_BACKEND``. For example:
 
-.. code-block :: shell
+.. code-block:: python
 
     AUTOGL_BACKEND=pyg python xxx.py
 
 or
 
-.. code-block :: python
+.. code-block:: python
 
     import os
     os.environ["AUTOGL_BACKEND"] = "pyg"