Update dependence conda package versions

.github/workflows/pythonpackage.yml

@@ -24,10 +24,10 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         pip install -e .
-        conda install pytorch torchvision cpuonly -c pytorch
-        conda install dgl -c dglteam
-        conda install pymatgen==2020.6.8 -c conda-forge
-        conda install rdkit==2020.03.3 -c conda-forge
+        conda install pytorch==1.6.0 torchvision cpuonly -c pytorch
+        conda install dgl==0.5.0 -c dglteam
+        conda install pymatgen==2020.8.13 -c conda-forge
+        conda install rdkit==2020.03.5 -c conda-forge
         conda install openbabel==3.1.1 -c conda-forge
     - name: Lint with flake8
       run: |

README.md

@@ -2,7 +2,7 @@ Use the pretrained model:
 [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/pretrained?filepath=bondnet%2Fscripts%2Fpredict_binder.ipynb)
 
 Train the model: 
-[![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/master?filepath=bondnet%2Fscripts%2Ftrain_bde.ipynb)
+[![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/pretrained?filepath=bondnet%2Fscripts%2Ftrain_bde.ipynb)
 
 
 # Table of Contents
@@ -76,7 +76,7 @@ can be found [here](./bondnet/scripts/examples/predict).
 
 The [train_bde.ipynb](./bondnet/scripts/train_bde.ipynb) Jupyter notebook shows
 how to train BonDNet on a BDE dataset of both neutral and charged molecules.
-Try it at: [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/master?filepath=bondnet%2Fscripts%2Ftrain_bde.ipynb)
+Try it at: [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/pretrained?filepath=bondnet%2Fscripts%2Ftrain_bde.ipynb)
 
 The [train_bde.ipynb](./bondnet/scripts/train_bde.ipynb) Jupyter notebook trains a model on CPU.
 If you want to train on GPUs (a single GPU or distributed), take a look at

binder/environment.yml

@@ -6,9 +6,9 @@ channels:
   - defaults
 dependencies:
   - pip
-  - pymatgen==2020.6.8
-  - rdkit==2020.03.3
+  - pymatgen==2020.8.13
+  - rdkit==2020.03.5
   - openbabel==3.1.1
-  - pytorch==1.5.0
+  - pytorch==1.6.0
   - torchvision
-  - dgl==0.4.3
+  - dgl==0.5.0

bondnet/data/grapher.py

@@ -83,27 +83,23 @@ def __init__(self, atom_featurizer=None, bond_featurizer=None, self_loop=True):
         )
 
     def build_graph(self, mol):
-        g = dgl.DGLGraph()
 
-        # Add nodes
         num_atoms = mol.GetNumAtoms()
-        g.add_nodes(num_atoms)
+        num_bonds = mol.GetNumBonds()
 
-        # Add edges
         src_list = []
         dst_list = []
-        num_bonds = mol.GetNumBonds()
         for i in range(num_bonds):
             bond = mol.GetBondWithIdx(i)
             u = bond.GetBeginAtomIdx()
             v = bond.GetEndAtomIdx()
             src_list.extend([u, v])
             dst_list.extend([v, u])
-        g.add_edges(src_list, dst_list)
-
         if self.self_loop:
-            nodes = g.nodes()
-            g.add_edges(nodes, nodes)
+            src_list.extend(range(num_atoms))
+            dst_list.extend(range(num_atoms))
+
+        g = dgl.graph((src_list, dst_list), num_nodes=num_atoms)
 
         # add name
         g.mol_name = mol.GetProp("_Name")
@@ -136,20 +132,16 @@ def __init__(self, atom_featurizer=None, bond_featurizer=None, self_loop=True):
 
     def build_graph(self, mol):
 
-        g = dgl.DGLGraph()
         num_atoms = mol.GetNumAtoms()
-        g.add_nodes(num_atoms)
 
         if self.self_loop:
-            g.add_edges(
-                [i for i in range(num_atoms) for j in range(num_atoms)],
-                [j for i in range(num_atoms) for j in range(num_atoms)],
-            )
+            src_list = [i for i in range(num_atoms) for j in range(num_atoms)]
+            dst_list = [j for i in range(num_atoms) for j in range(num_atoms)]
         else:
-            g.add_edges(
-                [i for i in range(num_atoms) for j in range(num_atoms - 1)],
-                [j for i in range(num_atoms) for j in range(num_atoms) if i != j],
-            )
+            src_list = [i for i in range(num_atoms) for j in range(num_atoms - 1)]
+            dst_list = [j for i in range(num_atoms) for j in range(num_atoms) if i != j]
+
+        g = dgl.graph((src_list, dst_list), num_nodes=num_atoms)
 
         # add name
         g.mol_name = mol.GetProp("_Name")

bondnet/scripts/examples/train/README.md

@@ -40,7 +40,7 @@ To train BonDNet, three files are needed:
 
 The [train_bde.ipynb](../../train_bde.ipynb) Jupyter notebook shows
 how to train BonDNet on a BDE dataset of both neutral and charged molecules.
-Try it at: [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/master?filepath=bondnet%2Fscripts%2Ftrain_bde.ipynb)
+Try it at: [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/mjwen/bondnet/pretrained?filepath=bondnet%2Fscripts%2Ftrain_bde.ipynb)
 
 The [train_bde.ipynb](../../train_bde.ipynb) Jupyter notebook trains a model on CPU.
 If you want to train on GPUs (a single GPU or distributed), take a look at

bondnet/scripts/train_bde.ipynb

@@ -6,7 +6,7 @@
    "source": [
     "# Train BonDNet \n",
     "\n",
-    "In this notebook, we show how to train the BonDNet graph neural network model for bond dissociation energy (BDE) prediction. We only show how to train on CPUs. See [train_bde_distributed.py]() for a script for training on GPUs (a single GPU or distributed training on multiple GPUs). "
+    "In this notebook, we show how to train the BonDNet graph neural network model for bond dissociation energy (BDE) prediction. We only show how to train on CPUs. See [train_bde_distributed.py](./) for a script for training on GPUs (a single GPU or distributed training on multiple GPUs). "
    ]
   },
   {
@@ -47,7 +47,7 @@
     "- `molecule_attributes.yaml` This file contains extra molecular attributes (charges here) for molecules given in `molecules.sdf`. Some molecular attributes can be inferred from its SDF block, and they are overrode by the attributes specified in the `molecule_attributes.yaml` file.  \n",
     "- `reactions.csv` This file list the bond dissociation reations formed by the molecules given in `molecules.sdf`. Each line lists the reactant, products, and BDE of a reaction. The reactant and products are specified by their index in `molecules.sdf`. \n",
     "\n",
-    "See [here]() for the three files used in this notebook. "
+    "See [here](./examples/train) for the three files used in this notebook. "
    ]
   },
   {

reminder.md

@@ -0,0 +1,17 @@
+This is a reminder file for the admin of the repo.
+
+# Update dependence conda package version
+
+When changing the version of dependence conda packages, the below files should be updated:
+- [README.md](./README.md)
+- [pythonpackage.yml](./.github/workflows/pythonpackage.yml)
+- [environment.yml](./binder/environment.yml)
+
+
+# Update binder link to Jupyter notebook
+
+When the Jupyter notebook used as demo on binder are replaced, the binder links in the
+below files should be updated. 
+- [README.md](./README.md)
+- [predict/README.md](./bondnet/scripts/examples/predict/README.md)
+- [train/README.md](./bondnet/scripts/examples/train/README.md)

tests/utils.py

@@ -172,11 +172,9 @@ def make_homo_CH2O():
     A global node u is attached to all atoms and bonds.
     """
 
-    g = dgl.DGLGraph()
-    g.add_nodes(4)
     src = [0, 1, 1, 1, 2, 3]
-    des = [1, 0, 2, 3, 1, 1]
-    g.add_edges(src, des)
+    dst = [1, 0, 2, 3, 1, 1]
+    g = dgl.graph((src, dst))
 
     feats = {}
     N = 4