Changes (bd0c2434) · Commits · 钟慕尧 / deepchem

deepchem/models/graph_models.py

+8 −8

Original line number	Diff line number	Diff line
		@@ -4,14 +4,14 @@ import deepchem as dc
		import numpy as np
		import tensorflow as tf

		from typing import List
		from deepchem.utils.typing import OneOrMany
		from typing import List, Union
		from deepchem.utils.typing import OneOrMany, KerasLossFn
		from deepchem.data import NumpyDataset, pad_features
		from deepchem.feat.graph_features import ConvMolFeaturizer
		from deepchem.feat.mol_graphs import ConvMol
		from deepchem.metrics import to_one_hot
		from deepchem.models import KerasModel, layers
		from deepchem.models.losses import L2Loss, SoftmaxCrossEntropy
		from deepchem.models.losses import L2Loss, SoftmaxCrossEntropy, Loss
		from deepchem.trans import undo_transforms
		from tensorflow.keras.layers import Input, Dense, Reshape, Softmax, Dropout, Activation, BatchNormalization

		@@ -72,7 +72,7 @@ class WeaveModel(KerasModel):
		n_atom_feat: OneOrMany[int] = 75,
		n_pair_feat: OneOrMany[int] = 14,
		n_hidden: int = 50,
		n_graph_feat: OneOrMany[int] = 128,
		n_graph_feat: int = 128,
		n_weave: int = 2,
		fully_connected_layer_sizes: List[int] = [2000, 1000],
		mode: str = "classification",
		@@ -161,7 +161,7 @@ class WeaveModel(KerasModel):
		output = Softmax()(logits)
		outputs = [output, logits]
		output_types = ['prediction', 'loss']
		loss = SoftmaxCrossEntropy()
		loss: Loss = SoftmaxCrossEntropy()
		else:
		output = Dense(n_tasks)(weave_gather)
		outputs = [output]
		@@ -774,7 +774,7 @@ class GraphConvModel(KerasModel):
		batch_size=batch_size)
		if mode == "classification":
		output_types = ['prediction', 'loss', 'embedding']
		loss = SoftmaxCrossEntropy()
		loss: Union[Loss, KerasLossFn] = SoftmaxCrossEntropy()
		else:
		if self.uncertainty:
		output_types = ['prediction', 'variance', 'loss', 'loss', 'embedding']

docs/models.rst

+69 −1

Original line number	Diff line number	Diff line
		@@ -144,9 +144,77 @@ XGBoostModel
		.. autoclass:: deepchem.models.XGBoostModel
		:members:


		Keras Models
		============
		DeepChem extensively uses `Keras`_ to build powerful machine learning models.

		Losses
		------

		.. autoclass:: deepchem.models.losses.Loss
		:members:

		.. autoclass:: deepchem.models.losses.L1Loss
		:members:

		.. autoclass:: deepchem.models.losses.L2Loss
		:members:

		.. autoclass:: deepchem.models.losses.HingeLoss
		:members:

		.. autoclass:: deepchem.models.losses.BinaryCrossEntropy
		:members:

		.. autoclass:: deepchem.models.losses.CategoricalCrossEntropy
		:members:

		.. autoclass:: deepchem.models.losses.SigmoidCrossEntropy
		:members:

		.. autoclass:: deepchem.models.losses.SoftmaxCrossEntropy
		:members:

		.. autoclass:: deepchem.models.losses.SparseSoftmaxCrossEntropy
		:members:

		.. autoclass:: deepchem.models.losses.SparseSoftmaxCrossEntropy
		:members:

		Optimizers
		----------

		.. autoclass:: deepchem.models.optimizers.Optimizer
		:members:

		.. autoclass:: deepchem.models.optimizers.LearningRateSchedule
		:members:

		.. autoclass:: deepchem.models.optimizers.Adam
		:members:

		.. autoclass:: deepchem.models.optimizers.RMSProp
		:members:

		.. autoclass:: deepchem.models.optimizers.GradientDescent
		:members:

		.. autoclass:: deepchem.models.optimizers.ExponentialDecay
		:members:

		.. autoclass:: deepchem.models.optimizers.PolynomialDecay
		:members:

		.. autoclass:: deepchem.models.optimizers.LinearCosineDecay
		:members:

		.. autoclass:: deepchem.models.optimizers.LinearCosineDecay
		:members:


		KerasModel
		----------
		DeepChem extensively uses `Keras`_ to build powerful machine learning models.

		Training loss and validation metrics can be automatically logged to `Weights & Biases`_ with the following commands::

Admin message