Update (#1447)

apps/life_sci/README.md

@@ -66,7 +66,7 @@ Once you have installed the package, you can verify the success of installation
 import dgllife
 
 print(dgllife.__version__)
-# 0.2.1
+# 0.2.2
 ```
 
 If you are new to DGL, the first time you import dgl a message will pop up as below:

apps/life_sci/conda/dgllife/meta.yaml

 package:
   name: dgllife{{ environ.get('APP_PACKAGE_SUFFIX', '') }}
-  version: "0.2.1"
+  version: "0.2.2"
 
 source:
-  url: https://files.pythonhosted.org/packages/7d/ca/543802eca0e8d263a73309379f79a054b3bb7e7377ead0b69d2451728465/dgllife-0.2.1.tar.gz
-  sha256: 11a1622ba587f7ff67f42ad8bc88d238faaa459ab554cf59e1b69a10079d41c0
+  url: https://files.pythonhosted.org/packages/19/e7/7d9890c7ddb303613f8a9c0bafafa40239e65f9c0698e699f19f577412fc/dgllife-0.2.2.tar.gz
+  sha256: 812a93d54cd5c049a7368aae01d53b5a4a12ca2d0e4e57d74fa4a844a1c01d92
 
 requirements:
   build:

apps/life_sci/examples/reaction_prediction/rexgen_direct/README.md

@@ -51,7 +51,7 @@ By default, we store the model per 10000 iterations in `center_results`.
 **Speedup**: For an epoch of training, our implementation takes about 5095s for the first epoch  while the authors' 
 implementation takes about 11657s, which is roughly a speedup by 2.3x.
 
-For model evaluation, we can choose whether to exclude reactants not contributing atoms to the product 
+For model evaluation, we can choose whether to exclude reactants not contributing heavy atoms to the product 
 (e.g. reagents and solvents) in top-k atom pair selection, which will make the task easier. 
 For the easier evaluation, do
 
@@ -64,8 +64,8 @@ A summary of the model performance is as follows:
 | Item            | Top 6 accuracy | Top 8 accuracy | Top 10 accuracy |
 | --------------- | -------------- | -------------- | --------------- |
 | Paper           | 89.8           | 92.0           | 93.3            |
-| Hard evaluation | 86.5           | 89.6           | 91.2            |
-| Easy evaluation | 88.9           | 92.0           | 93.5            |
+| Hard evaluation | 88.8           | 91.6           | 92.9            |
+| Easy evaluation | 91.0           | 93.7           | 94.9            |
 
 ### Pre-trained Model
 

apps/life_sci/python/dgllife/data/csv_dataset.py

@@ -31,7 +31,7 @@ class MoleculeCSVDataset(object):
         Featurization for edges like bonds in a molecule, which can be used to update
         edata for a DGLGraph.
     smiles_column: str
-        Column name that including smiles in ``df``.
+        Column name for smiles in ``df``.
     cache_file_path: str
         Path to store the preprocessed DGLGraphs. For example, this can be ``'dglgraph.bin'``.
     task_names : list of str or None

apps/life_sci/python/dgllife/libinfo.py

 """Information for the library."""
 
 # current version
-__version__ = '0.2.1'
+__version__ = '0.2.2'

apps/life_sci/python/dgllife/model/gnn/wln.py

@@ -141,6 +141,8 @@ class WLN(nn.Module):
             concat_edge_feats = torch.cat([g.edata['he_src'], edge_feats], dim=1)
             g.edata['he'] = self.project_concatenated_messages(concat_edge_feats)
             g.update_all(fn.copy_edge('he', 'm'), fn.sum('m', 'hv_new'))
+            node_feats = self.get_new_node_feats(
+                torch.cat([node_feats, g.ndata['hv_new']], dim=1))
 
         g = g.local_var()
         g.ndata['hv'] = self.project_node_messages(node_feats)

apps/life_sci/python/dgllife/model/pretrain.py

@@ -22,7 +22,7 @@ URL = {
     'DGMG_ZINC_canonical': 'pre_trained/dgmg_ZINC_canonical.pth',
     'DGMG_ZINC_random': 'pre_trained/dgmg_ZINC_random.pth',
     'JTNN_ZINC': 'pre_trained/JTNN_ZINC.pth',
-    'wln_center_uspto': 'dgllife/pre_trained/wln_center_uspto.pth'
+    'wln_center_uspto': 'dgllife/pre_trained/wln_center_uspto_v2.pth'
 }
 
 def download_and_load_checkpoint(model_name, model, model_postfix,

apps/life_sci/python/update_version.py

@@ -8,7 +8,7 @@ List of affected files:
 import os
 import re
 
-__version__ = "0.2.1"
+__version__ = "0.2.2"
 print(__version__)
 
 # Implementations