Merge pull request #425 from corochann/refactor_wle

Refactor wle

chainer_chemistry/dataset/parsers/data_frame_parser.py

@@ -162,7 +162,8 @@ def parse(self, df, return_smiles=False, target_index=None,
         else:
             raise NotImplementedError
 
-        smileses = numpy.array(smiles_list) if return_smiles else None
+        smileses = numpy.array(
+            smiles_list, dtype=object) if return_smiles else None
         if return_is_successful:
             is_successful = numpy.array(is_successful_list)
         else:

chainer_chemistry/dataset/parsers/sdf_file_parser.py

@@ -149,7 +149,8 @@ def parse(self, filepath, return_smiles=False, target_index=None,
             # Spec not finalized yet for general case
             result = pp.process(filepath)
 
-        smileses = numpy.array(smiles_list) if return_smiles else None
+        smileses = numpy.array(
+            smiles_list, dtype=object) if return_smiles else None
         if return_is_successful:
             is_successful = numpy.array(is_successful_list)
         else:

chainer_chemistry/datasets/molnet/molnet.py

@@ -113,12 +113,7 @@ def postprocess_label(label_list):
             raise TypeError("split must be None, str or instance of"
                             " BaseSplitter, but got {}".format(type(split)))
 
-        if isinstance(splitter, ScaffoldSplitter):
-            get_smiles = True
-        else:
-            get_smiles = return_smiles
-
-        if isinstance(splitter, DeepChemScaffoldSplitter):
+        if isinstance(splitter, (ScaffoldSplitter, DeepChemScaffoldSplitter)):
             get_smiles = True
         else:
             get_smiles = return_smiles

examples/molnet_wle/README.md

@@ -4,6 +4,16 @@ In this directory, we provide an implementaion of [Weisfeiler-Lehman Embedding (
 
 ## How to run the code
 
+### Test run command
+
+```bash
+# Training tox21 dataset using RSGCN-CWLE model. Short 3 epoch for testing.
+python train_molnet_wle.py --dataset tox21 --method rsgcn_cwle --epoch 3  --device 0
+
+# Prediction with trained model
+python predict_molnet_wle.py --dataset tox21 --method rsgcn_cwle --in-dir result --device 0
+```
+
 ### Train the model by specifying dataset
 
 Basically, no changes from the original molnet examples (examples/molnet/train_molnet.py).
@@ -14,7 +24,7 @@ To test WLE, choose one of 'xxx_wle', 'xxx_cwle', and 'xxx_gwle' where 'xxx' is
 - xxx_cwle (recommended): apply the Concat WLE to the GNN 'xxx'
 - xxx_gwle: apply the Gated-sum WLE to the GNN 'xxx'
 
-#### additional options
+#### Additional options
 
 Introducing the WLE, we have some more additional options.
 In general you do not need to specify these options (use the default values!).

examples/molnet_wle/predict_molnet_wle.py

@@ -2,14 +2,12 @@
 from __future__ import print_function
 
 import argparse
-import json
 import os
 
 import chainer
 from chainer.iterators import SerialIterator
 from chainer.training.extensions import Evaluator
 from chainer_chemistry.training.extensions.roc_auc_evaluator import ROCAUCEvaluator  # NOQA
-from chainer import cuda
 # Proposed by Ishiguro
 # ToDo: consider go/no-go with following modification
 # Re-load the best-validation score snapshot using serializers

examples/molnet_wle/train_molnet_wle.py

@@ -16,8 +16,6 @@
 from chainer.training import extensions as E
 
 from chainer_chemistry.dataset.converters import converter_method_dict
-from chainer_chemistry.config import MAX_ATOMIC_NUM
-from chainer_chemistry.dataset.converters import concat_mols
 from chainer_chemistry.dataset.preprocessors import preprocess_method_dict, wle
 from chainer_chemistry import datasets as D
 from chainer_chemistry.datasets.molnet.molnet_config import molnet_default_config  # NOQA
@@ -27,7 +25,6 @@
 from chainer_chemistry.models.prediction import Classifier
 from chainer_chemistry.models.prediction import Regressor
 from chainer_chemistry.models.prediction import set_up_predictor
-from chainer_chemistry.training.extensions import BatchEvaluator, ROCAUCEvaluator  # NOQA
 from chainer_chemistry.training.extensions.auto_print_report import AutoPrintReport  # NOQA
 from chainer_chemistry.utils import save_json
 from chainer_chemistry.models.cwle.cwle_graph_conv_model import MAX_WLE_NUM
@@ -156,11 +153,7 @@ def download_entire_dataset(dataset_name, num_data, labels, method, cache_dir, a
                                                 labels=labels,
                                                 split=dc_scaffold_splitter,
                                                 target_index=target_index)
-    # To use the splitter defined in the config file
-    #dataset_parts = D.molnet.get_molnet_dataset(dataset_name, preprocessor,
-    #                                            labels=labels,
-    #                                            target_index=target_index)
-
+
     dataset_parts = dataset_parts['dataset']
 
     # Cache the downloaded dataset.