Reorganized MolNet materials loaders (dec031e0) · Commits · 钟慕尧 / deepchem

deepchem/feat/materials_featurizers.py

+5 −1

Original line number	Diff line number	Diff line
		@@ -37,6 +37,10 @@ class ElementPropertyFingerprint(MaterialCompositionFeaturizer):

		.. [4] Pymatgen: Ong, S.P. et al. Comput. Mater. Sci. 68, 314-319 (2013).

		Notes
		-----
		`NaN` feature values are automatically converted to 0 by this featurizer.

		"""

		def __init__(self, data_source='matminer'):

deepchem/molnet/init.py

+2 −1

Original line number	Diff line number	Diff line
		@@ -31,7 +31,8 @@ from deepchem.molnet.load_function.kinase_datasets import load_kinase
		from deepchem.molnet.load_function.thermosol_datasets import load_thermosol
		from deepchem.molnet.load_function.hppb_datasets import load_hppb
		from deepchem.molnet.load_function.chembl25_datasets import load_chembl25
		from deepchem.molnet.load_function.material_datasets import load_bandgap, load_perovskite
		from deepchem.molnet.load_function.material_datasets.load_bandgap import load_bandgap
		from deepchem.molnet.load_function.material_datasets.load_perovskite import load_perovskite

		from deepchem.molnet.dnasim import simulate_motif_density_localization
		from deepchem.molnet.dnasim import simulate_motif_counting

deepchem/molnet/load_function/material_datasets.py

+20 −7

Original line number	Diff line number	Diff line
		@@ -4,12 +4,12 @@ Datasets for inorganic crystal structures.
		import os
		import logging
		import deepchem
		from deepchem.feat import Featurizer
		from deepchem.feat import Featurizer, MaterialStructureFeaturizer, MaterialCompositionFeaturizer
		from deepchem.trans import Transformer
		from deepchem.splits.splitters import Splitter
		from deepchem.molnet.defaults import get_defaults

		from typing import List, Tuple, Dict, Optional
		from typing import List, Tuple, Dict, Optional, Union

		logger = logging.getLogger(__name__)

		@@ -41,7 +41,7 @@ DEFAULT_SPLITTERS = {k: DEFAULT_SPLITTERS[k] for k in splitters}


		def load_bandgap(
		featurizer: Featurizer = DEFAULT_FEATURIZERS['ElementPropertyFingerprint'],
		featurizer: MaterialCompositionFeaturizer = DEFAULT_FEATURIZERS['ElementPropertyFingerprint'],
		transformers: List[Transformer] = [
		DEFAULT_TRANSFORMERS['NormalizationTransformer']
		],
		@@ -64,15 +64,21 @@ def load_bandgap(
		"""Load band gap dataset.

		Contains 4604 experimentally measured band gaps for inorganic
		crystal structure compositions.
		crystal structure compositions. In benchmark studies, random forest
		models achieved a mean average error of 0.45 eV during five-fold
		nested cross validation on this dataset.

		For more details on the dataset see [1]_. For more details
		on previous benchmarks for this dataset, see [2]_.

		Parameters
		----------
		featurizer : ElementPropertyFingerprint
		featurizer : MaterialCompositionFeaturizer
		(default ElementPropertyFingerprint)
		A featurizer that inherits from deepchem.feat.Featurizer.
		transformers : List{List of allowed transformers for this dataset}
		transformers : List[Transformer]
		A transformer that inherits from deepchem.trans.Transformer.
		splitter : RandomSplitter
		splitter : Splitter (default RandomSplitter)
		A splitter that inherits from deepchem.splits.splitters.Splitter.
		reload : bool (default True)
		Try to reload dataset from disk if already downloaded. Save to disk
		@@ -224,6 +230,13 @@ def load_perovskite(
		"""Load perovskite dataset.

		Contains 18928 perovskite structures and their formation energies.
		In benchmark studies, random forest models and crystal graph
		neural networks achieved mean average error of 0.23 and 0.05 eV/atom,
		respectively, during five-fold nested cross validation on this
		dataset.

		For more details on the dataset see [1]_. For more details
		on previous benchmarks for this dataset, see [2]_.

		Parameters
		----------

deepchem/molnet/load_function/material_datasets/init.py

0 → 100644

+0 −0

Empty file added.

deepchem/molnet/load_function/material_datasets/load_bandgap.py

0 → 100644

+205 −0

Original line number	Diff line number	Diff line
		"""
		Experimental bandgaps for inorganic crystals.
		"""
		import os
		import logging
		import deepchem
		from deepchem.feat import Featurizer, MaterialStructureFeaturizer, MaterialCompositionFeaturizer
		from deepchem.trans import Transformer
		from deepchem.splits.splitters import Splitter
		from deepchem.molnet.defaults import get_defaults

		from typing import List, Tuple, Dict, Optional, Union

		logger = logging.getLogger(__name__)

		# TODO: Change URLs
		DEFAULT_DIR = deepchem.utils.get_data_dir()
		BANDGAP_URL = 'http://deepchem.io.s3-website-us-west-1.amazonaws.com/datasets/expt_gap.tar.gz'

		# dict of accepted featurizers for this dataset
		# modify the returned dicts for your dataset
		DEFAULT_FEATURIZERS = get_defaults("feat")

		# Names of supported featurizers
		featurizers = [
		'ElementPropertyFingerprint',
		]
		DEFAULT_FEATURIZERS = {k: DEFAULT_FEATURIZERS[k] for k in featurizers}

		# dict of accepted transformers
		DEFAULT_TRANSFORMERS = get_defaults("trans")

		# dict of accepted splitters
		DEFAULT_SPLITTERS = get_defaults("splits")

		# names of supported splitters
		splitters = ['RandomSplitter']
		DEFAULT_SPLITTERS = {k: DEFAULT_SPLITTERS[k] for k in splitters}


		def load_bandgap(
		featurizer: MaterialCompositionFeaturizer = DEFAULT_FEATURIZERS[
		'ElementPropertyFingerprint'],
		transformers: List[Transformer] = [
		DEFAULT_TRANSFORMERS['NormalizationTransformer']
		],
		splitter: Splitter = DEFAULT_SPLITTERS['RandomSplitter'],
		reload: bool = True,
		data_dir: Optional[str] = None,
		save_dir: Optional[str] = None,
		featurizer_kwargs: Dict[str, object] = {'data_source': 'matminer'},
		splitter_kwargs: Dict[str, object] = {
		'frac_train': 0.8,
		'frac_valid': 0.1,
		'frac_test': 0.1
		},
		transformer_kwargs: Dict[str, Dict[str, object]] = {
		'NormalizationTransformer': {
		'transform_X': True
		}
		},
		**kwargs) -> Tuple[List, Tuple, List]:
		"""Load band gap dataset.

		Contains 4604 experimentally measured band gaps for inorganic
		crystal structure compositions. In benchmark studies, random forest
		models achieved a mean average error of 0.45 eV during five-fold
		nested cross validation on this dataset.

		For more details on the dataset see [1]_. For more details
		on previous benchmarks for this dataset, see [2]_.

		Parameters
		----------
		featurizer : MaterialCompositionFeaturizer
		(default ElementPropertyFingerprint)
		A featurizer that inherits from deepchem.feat.Featurizer.
		transformers : List[Transformer]
		A transformer that inherits from deepchem.trans.Transformer.
		splitter : Splitter (default RandomSplitter)
		A splitter that inherits from deepchem.splits.splitters.Splitter.
		reload : bool (default True)
		Try to reload dataset from disk if already downloaded. Save to disk
		after featurizing.
		data_dir : str, optional
		Path to datasets.
		save_dir : str, optional
		Path to featurized datasets.
		featurizer_kwargs : dict
		Specify parameters to featurizer, e.g. {"size": 1024}
		splitter_kwargs : dict
		Specify parameters to splitter, e.g. {"seed": 42}
		transformer_kwargs : dict
		Maps transformer names to constructor arguments, e.g.
		{"BalancingTransformer": {"transform_x":True, "transform_y":False}}
		**kwargs : additional optional arguments.

		Returns
		-------
		tasks, datasets, transformers : tuple
		tasks : list
		Column names corresponding to machine learning target variables.
		datasets : tuple
		train, validation, test splits of data as
		``deepchem.data.datasets.Dataset`` instances.
		transformers : list
		``deepchem.trans.transformers.Transformer`` instances applied
		to dataset.

		References
		----------
		.. [1] Zhuo, Y. et al. "Predicting the Band Gaps of Inorganic Solids by Machine Learning." J. Phys. Chem. Lett. (2018) DOI: 10.1021/acs.jpclett.8b00124.

		.. [2] Dunn, A. et al. "Benchmarking Materials Property Prediction Methods: The Matbench Test Set and Automatminer Reference Algorithm." https://arxiv.org/abs/2005.00707 (2020)

		Examples
		--------
		>> import deepchem as dc
		>> tasks, datasets, transformers = dc.molnet.load_bandgap(reload=False)
		>> train_dataset, val_dataset, test_dataset = datasets
		>> n_tasks = len(tasks)
		>> n_features = train_dataset.get_data_shape()[0]
		>> model = dc.models.MultitaskRegressor(n_tasks, n_features)

		"""

		# Featurize
		logger.info("About to featurize band gap dataset.")
		my_tasks = ['experimental_bandgap'] # machine learning targets

		# Get DeepChem data directory if needed
		if data_dir is None:
		data_dir = DEFAULT_DIR
		if save_dir is None:
		save_dir = DEFAULT_DIR

		if issubclass(featurizer, MaterialCompositionFeaturizer):
		featurizer = featurizer(**featurizer_kwargs)
		else:
		raise TypeError(
		"featurizer must be a subclass of MaterialCompositionFeaturizer.")

		if issubclass(splitter, Splitter):
		splitter = splitter()
		else:
		raise TypeError("splitter must be a subclass of Splitter.")

		# Reload from disk
		if reload:
		featurizer_name = str(featurizer.__class__.__name__)
		splitter_name = str(splitter.__class__.__name__)
		save_folder = os.path.join(save_dir, "bandgap-featurized", featurizer_name,
		splitter_name)

		loaded, all_dataset, transformers = deepchem.utils.save.load_dataset_from_disk(
		save_folder)
		if loaded:
		return my_tasks, all_dataset, transformers

		# First type of supported featurizers
		supported_featurizers = ['ElementPropertyFingerprint'
		] # type: List[Featurizer]

		# Load .tar.gz file
		if featurizer.__class__.__name__ in supported_featurizers:
		dataset_file = os.path.join(data_dir, 'expt_gap.tar.gz')
		deepchem.utils.untargz_file(dataset_file, dest_dir=data_dir)
		dataset_file = os.path.join(data_dir, 'expt_gap.json')

		if not os.path.exists(dataset_file):
		deepchem.utils.download_url(url=BANDGAP_URL, dest_dir=data_dir)
		deepchem.utils.untargz_file(
		os.path.join(data_dir, 'expt_gap.tar.gz'), data_dir)

		# Changer loader to match featurizer and data file type
		loader = deepchem.data.JsonLoader(
		tasks=my_tasks,
		feature_field="composition",
		label_field="experimental_bandgap",
		featurizer=featurizer)

		# Featurize dataset
		dataset = loader.create_dataset(dataset_file)

		train_dataset, valid_dataset, test_dataset = splitter.train_valid_test_split(
		dataset, **splitter_kwargs)

		# Initialize transformers
		transformers = [
		DEFAULT_TRANSFORMERS[t](dataset=dataset, **transformer_kwargs[t])
		if isinstance(t, str) else t(
		dataset=dataset, **transformer_kwargs[str(t.__name__)])
		for t in transformers
		]

		for transformer in transformers:
		train_dataset = transformer.transform(train_dataset)
		valid_dataset = transformer.transform(valid_dataset)
		test_dataset = transformer.transform(test_dataset)

		if reload: # save to disk
		deepchem.utils.save.save_dataset_to_disk(
		save_folder, train_dataset, valid_dataset, test_dataset, transformers)

		return my_tasks, (train_dataset, valid_dataset, test_dataset), transformers

Admin message