2 class case to sigmoidloss, stratified split (5ed8b888) · Commits · 钟慕尧 / deepchem

deepchem/models/tensorgraph/models/chemnet_models.py

+18 −10

Original line number	Diff line number	Diff line
		@@ -17,11 +17,11 @@ import logging

		from deepchem.data.datasets import pad_batch
		from deepchem.models import KerasModel, layers
		from deepchem.models.losses import L2Loss, SoftmaxCrossEntropy
		from deepchem.models.losses import L2Loss, SoftmaxCrossEntropy, SigmoidCrossEntropy
		from deepchem.metrics import to_one_hot
		from deepchem.models.tensorgraph.layers import KerasLayer
		from deepchem.models.tensorgraph import chemnet_layers
		from tensorflow.keras.layers import Input, Dense, Reshape, Softmax
		from tensorflow.keras.layers import Input, Dense, Reshape, Softmax, Activation
		from tensorflow.keras.layers import Dropout, Conv1D, Concatenate, Lambda, GRU, LSTM, Bidirectional
		from tensorflow.keras.layers import Conv2D, ReLU, Add, GlobalAveragePooling2D

		@@ -154,12 +154,16 @@ class Smiles2Vec(KerasModel):
		rnn_embeddings = layer(rnn_embeddings)

		if self.mode == "classification":
		logits = Dense(self.n_tasks * 2)(rnn_embeddings)
		logits = Reshape((self.n_tasks, 2))(logits)
		logits = Dense(self.n_tasks * self.n_classes)(rnn_embeddings)
		logits = Reshape((self.n_tasks, self.n_classes))(logits)
		if self.n_classes == 2:
		output = Activation(activation='sigmoid')(logits)
		loss = SigmoidCrossEntropy()
		else:
		output = Softmax()(logits)
		loss = SoftmaxCrossEntropy()
		outputs = [output, logits]
		output_types = ['prediction', 'loss']
		loss = SoftmaxCrossEntropy()

		else:
		output = Dense(self.n_tasks * 1, name='Dense')(rnn_embeddings)
		@@ -276,12 +280,16 @@ class ChemCeption(KerasModel):
		avg_pooling_out = GlobalAveragePooling2D()(inceptionC_out)

		if self.mode == "classification":
		logits = Dense(self.n_tasks * 2)(avg_pooling_out)
		logits = Reshape((self.n_tasks, 2))(logits)
		logits = Dense(self.n_tasks * self.n_classes)(rnn_embeddings)
		logits = Reshape((self.n_tasks, self.n_classes))(logits)
		if self.n_classes == 2:
		output = Activation(activation='sigmoid')(logits)
		loss = SigmoidCrossEntropy()
		else:
		output = Softmax()(logits)
		loss = SoftmaxCrossEntropy()
		outputs = [output, logits]
		output_types = ['prediction', 'loss']
		loss = SoftmaxCrossEntropy()

		else:
		output = Dense(self.n_tasks * 1)(avg_pooling_out)

deepchem/molnet/load_function/sampl_datasets.py

+3 −1

Original line number	Diff line number	Diff line
		@@ -85,8 +85,10 @@ def load_sampl(featurizer='ECFP',
		splitters = {
		'index': deepchem.splits.IndexSplitter(),
		'random': deepchem.splits.RandomSplitter(),
		'scaffold': deepchem.splits.ScaffoldSplitter()
		'scaffold': deepchem.splits.ScaffoldSplitter(),
		'stratified': deepchem.splits.SingletaskStratifiedSplitter(task_number=0)
		}

		splitter = splitters[split]
		logger.info("About to split dataset with {} splitter.".format(split))
		frac_train = kwargs.get("frac_train", 0.8)

Admin message