Merge pull request #15 from kelizhang/master

2021.5.31: Adjustment for the project structure and code optimizations.

gcastle/README.cn.md

@@ -92,4 +92,4 @@ print(mt.metrics)
 * 算法库拓展：新增`GES`,`HPCI`,`TTPM`等因果结构学习算法，并提供简易的可配置脚本辅助进行相应算法的调用和运行。
 * 真实场景数据集：将陆续公开一批来源于真实AIOPS场景的时间序列和事件序列数据集，其中真实的因果图标注来源于业务专家经验。
 
-欢迎大家使用`gCastle`. 该项目尚处于起步阶段，欢迎各个经验等级的贡献者。有任何疑问及建议，包括修改bug、贡献算法、完善文档等，请在社区提交issue，我们会及时回复交流。
+欢迎大家使用`gCastle`. 该项目尚处于起步阶段，欢迎各个经验等级的贡献者。有任何疑问及建议，包括修改bug、贡献算法、完善文档等，请在社区提交issue，我们会及时回复交流。

gcastle/castle/algorithms/__init__.py

@@ -22,7 +22,7 @@
 from .gradient import NotearsSob
 from .gradient import NotearsLowRank
 from .gradient import GOLEM
-from .gradient import GraN_DAG, Parameters
+from .gradient import GraN_DAG
 from .gradient import GAE
 from .gradient import MCSL
 from .gradient import RL

gcastle/castle/algorithms/gradient/__init__.py

@@ -19,7 +19,7 @@
 from .notears.low_rank import NotearsLowRank
 from .notears.golem import GOLEM
 
-from .gran_dag.gran_dag import GraN_DAG, Parameters
+from .gran_dag.gran_dag import GraN_DAG
 from .graph_auto_encoder.gae import GAE
 from .masked_csl.mcsl import MCSL
 

gcastle/castle/algorithms/gradient/corl1/corl1.py

@@ -51,7 +51,7 @@ class CORL1(BaseLearner):
     >>> from castle.metrics import MetricsDAG
     >>> true_dag, X = load_dataset(name='iid_test')
     >>> n = CORL1()
-    >>> n.learn(X, dag=true_dag)
+    >>> n.learn(X)
     >>> GraphDAG(n.causal_matrix, true_dag)
     >>> met = MetricsDAG(n.causal_matrix, true_dag)
     >>> print(met.metrics)

gcastle/castle/algorithms/gradient/corl2/corl2.py

@@ -50,7 +50,7 @@ class CORL2(BaseLearner):
     >>> from castle.metrics import MetricsDAG
     >>> true_dag, X = load_dataset(name='iid_test')
     >>> n = CORL2()
-    >>> n.learn(X, dag=true_dag)
+    >>> n.learn(X)
     >>> GraphDAG(n.causal_matrix, true_dag)
     >>> met = MetricsDAG(n.causal_matrix, true_dag)
     >>> print(met.metrics)

gcastle/castle/algorithms/gradient/gran_dag/__init__.py

@@ -13,4 +13,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .gran_dag import GraN_DAG, Parameters
+from .gran_dag import GraN_DAG

gcastle/castle/algorithms/gradient/gran_dag/base/__init__.py

@@ -1,5 +1,3 @@
-from .base_model import BaseModel
-from .base_model import LearnableModel
 from .base_model import NonlinearGauss
 from .base_model import NonlinearGaussANM
 

gcastle/castle/algorithms/gradient/gran_dag/base/base_model.py

@@ -13,9 +13,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import os
-import pickle
-
 import numpy as np
 import torch
 import torch.nn as nn
@@ -86,7 +83,6 @@ def forward_given_params(self, x, weights, biases):
             the parameters of each variable conditional
         """
 
-        num_zero_weights = 0
         for k in range(self.hidden_num + 1):
             # apply affine operator
             if k == 0:
@@ -95,18 +91,13 @@ def forward_given_params(self, x, weights, biases):
             else:
                 x = torch.einsum("tij,btj->bti", weights[k], x) + biases[k]
 
-            # count num of zeros
-            num_zero_weights += weights[k].numel() - torch.nonzero(weights[k]).size()[0]
-
             # apply non-linearity
             if k != self.hidden_num:
                 if self.nonlinear == "leaky-relu":
                     x = F.leaky_relu(x)
                 else:
                     x = torch.sigmoid(x)
 
-        self.zero_weights_ratio = num_zero_weights / float(self.numel_weights)
-
         return torch.unbind(x, 1)
 
     def get_w_adj(self):
@@ -218,26 +209,6 @@ def get_grad_norm(self, mode="wbx"):
 
         return torch.sqrt(grad_norm)
 
-    def save_parameters(self, exp_path, mode="wbx"):
-        """
-
-        Parameters
-        ----------
-        exp_path : str
-            path for saving model parameters
-        mode : str
-            w=weights, b=biases, x=extra_params (order is irrelevant)
-        """
-        params = self.get_parameters(mode=mode)
-        # save
-        with open(os.path.join(exp_path, "params_"+mode), 'wb') as f:
-            pickle.dump(params, f)
-
-    def load_parameters(self, exp_path, mode="wbx"):
-        with open(os.path.join(exp_path, "params_"+mode), 'rb') as f:
-            params = pickle.load(f)
-        self.set_parameters(params, mode=mode)
-
     def get_distribution(self, density_params):
         raise NotImplementedError