Merge pull request #26 from leynier/megexplainer-migration

meg explainer migration

config/leynier/meg-test-bbbp.json

@@ -0,0 +1,78 @@
+{
+    "experiment": {
+        "scope": "meg_Test",
+        "parameters": {
+            "lock_release_tout": 120,
+            "propagate": []
+        }
+    },
+    "do-pairs": [
+        {
+            "dataset": {
+                "class": "src.dataset.dataset_base.Dataset",
+                "parameters": {
+                    "generator": {
+                        "class": "src.dataset.generators.bbbp.BBBP",
+                        "parameters": {
+                            "data_dir": "data/datasets/bbbp/",
+                            "data_file_name": "BBBP-reduced.csv",
+                            "data_label_name": "p_np"
+                        }
+                    },
+                    "n_splits": 2
+                }
+            },
+            "oracle": {
+                "class": "src.oracle.nn.torch.OracleTorch",
+                "parameters": {
+                    "epochs": 50,
+                    "batch_size": 2,
+                    "optimizer": {
+                        "class": "torch.optim.RMSprop",
+                        "parameters": {
+                            "lr": 0.001,
+                            "momentum": 0.5
+                        }
+                    },
+                    "loss_fn": {
+                        "class": "torch.nn.CrossEntropyLoss",
+                        "parameters": {
+                            "reduction": "mean"
+                        }
+                    },
+                    "model": {
+                        "class": "src.oracle.nn.gcn.DownstreamGCN",
+                        "parameters": {
+                            "num_conv_layers": 2,
+                            "num_dense_layers": 1,
+                            "conv_booster": 2,
+                            "linear_decay": 1.8
+                        }
+                    }
+                }
+            }
+        }
+    ],
+    "explainers": [
+        {
+            "class": "src.explainer.rl.meg.MEGExplainer",
+            "parameters": {
+                "num_input": 1024,
+                "env": {
+                    "class": "src.explainer.rl.meg_utils.environments.bbbp_env.BBBPEnvironment",
+                    "parameters": {}
+                },
+                "action_encoder": {
+                    "class": "src.explainer.rl.meg_utils.utils.encoders.MorganCountFingerprintActionEncoder",
+                    "parameters": {}
+                }
+            }
+        },
+        {
+            "class": "src.explainer.search.dces.DCESExplainer",
+            "parameters": {}
+        }
+    ],
+    "compose_mes": "config/snippets/default_metrics.json",
+    "compose_strs": "config/snippets/default_store_paths.json"
+}

data/datasets/bbbp/BBBP-reduced.csv

@@ -0,0 +1,5 @@
+num,name,p_np,smiles
+1,Propanolol,1,[Cl].CC(C)NCC(O)COc1cccc2ccccc12
+2,Terbutylchlorambucil,1,C(=O)(OC(C)(C)C)CCCc1ccc(cc1)N(CCCl)CCCl
+3,SKF-93619,0,c1cc2c(cc(CC3=CNC(=NC3=O)NCCSCc3oc(cc3)CN(C)C)cc2)cc1
+4,M2L-663581,0,OCC(C)(O)c1onc(c2ncn3c2CN(C)C(c4c3cccc4Cl)=O)n1

main.py

@@ -19,7 +19,21 @@
 from src.utils.context import Context
 import sys
 
+try:
+    from dotenv import load_dotenv
+
+    load_dotenv()
+except Exception:
+    pass
+
 if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        # If no arguments are passed, try to find GRETEL_CONFIG_FILE in the environment
+        if "GRETEL_CONFIG_FILE" in os.environ:
+            sys.argv.append(os.environ["GRETEL_CONFIG_FILE"])
+        else:
+            print("Usage: python main.py <config_file> [run_number]")
+            sys.exit(1)
     print(f"Generating context for: {sys.argv[1]}")
     context = Context.get_context(sys.argv[1])
     context.run_number = int(sys.argv[2]) if len(sys.argv) == 3 else -1

src/dataset/generators/mol_gen.py

@@ -2,13 +2,15 @@
 
 import numpy as np
 import pandas as pd
-import networkx as nx
 from rdkit import Chem
 from rdkit.Chem import MolFromSmiles as smi2mol
 from rdkit.Chem import MolToSmiles as mol2smi
 
 from src.dataset.generators.base import Generator
-from src.dataset.instances.graph import GraphInstance
+from src.explainer.rl.meg_utils.utils.molecular_instance import (
+    MolecularInstance,
+)
+
 
 class MolGenerator(Generator):
 
@@ -65,7 +67,7 @@ def smile2graph(id, smile, label, dataset):
     g = None
     if sanitized:
         A,X,W = mol_to_matrices(mol, dataset)
-        g = GraphInstance(id=id, 
+        g = MolecularInstance(id=id, 
                         label=int(label), 
                         data=A, 
                         node_features=X, 
@@ -149,4 +151,4 @@ def rdk_enum_type_to_map(enum_type,offset=0):
 def rdk_enum_val_to_one_hot(enum_type):
     one_hot_vec = np.zeros((1,len(enum_type.values)))
     one_hot_vec[0,enum_type.real]=1
-    return one_hot_vec
+    return one_hot_vec

src/dataset/instances/base.py

@@ -1,12 +1,13 @@
+from typing import Optional
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from src.dataset.dataset_base import Dataset
+
+
 class DataInstance:
-
-    def __init__(self, id, label, data, dataset=None):
+    def __init__(self, id, label, data, dataset: Optional["Dataset"] = None):
         self.id = id
         self.data = data
-        self.label = label #TODO: Refactoring to have a one-hot encoding of labels!
+        self.label = label # TODO: Refactoring to have a one-hot encoding of labels!
         self._dataset = dataset
-
-
-
-
-

src/dataset/instances/graph.py

@@ -17,7 +17,19 @@ def __init__(self, id, label, data, node_features=None, edge_features=None, edge
         self.graph_features = graph_features
         self._nx_repr = None
 
+        num_nodes = self.data.shape[0]
+        num_edges = np.count_nonzero(self.data)
+        assert len(self.node_features) == num_nodes
+        assert len(self.edge_features) == num_edges
+        assert len(self.edge_weights) == num_edges
+
     def __deepcopy__(self, memo):
+        num_nodes = self.data.shape[0]
+        num_edges = np.count_nonzero(self.data)
+        assert len(self.node_features) == num_nodes
+        assert len(self.edge_features) == num_edges
+        assert len(self.edge_weights) == num_edges
+
         # Fields that are being shallow copied
         _dataset = self._dataset
 
@@ -29,7 +41,7 @@ def __deepcopy__(self, memo):
         _edge_features = deepcopy(self.edge_features, memo)
         _edge_weights = deepcopy(self.edge_weights, memo)
         _graph_features = deepcopy(self.graph_features, memo)
-        return GraphInstance(_new_id, _new_label, _data, _node_features, _edge_features, _edge_weights, _graph_features)
+        return GraphInstance(_new_id, _new_label, _data, _node_features, _edge_features, _edge_weights, _graph_features, _dataset)
 
     def get_nx(self):
         if not self._nx_repr:
@@ -73,4 +85,4 @@ def degree(self,node):
         return len(self.neighbors(node))
 
     def degrees(self):
-        return [ len(self.neighbors(y)) for y in self.nodes()]
+        return [ len(self.neighbors(y)) for y in self.nodes()]