def process(self, **kwargs):
"""This method is used for process your inputs, which accepts
only keyword arguments in your defined method 'process_step'.
This method will process the inputs, and transform them into tensors.
Commonly used keyword arguments:
--------------------------------
adj_transform: string, Callable function, or a tuple with function and dict arguments.
transform for adjacency matrix.
attr_transform: string, Callable function, or a tuple with function and dict arguments.
transform for attribute matrix.
graph_transform: string, Callable function, or a tuple with function and dict arguments.
transform for the entire graph, it is used before 'adj_transform' and 'attr_transform'.
other arguments (if have) will be passed into your method 'process_step'.
"""
cfg = self.cfg.process
_, kwargs = gf.wrapper(self.process_step)(**kwargs)
cfg.merge_from_dict(kwargs)
for k, v in cfg.items():
if k.endswith("transform"):
setattr(self.transform, k, gf.get(v))
self.is_processed = True
return self
def make_data(self, graph, graph_transform=None, device=None, **kwargs):
"""This method is used for process your inputs, which accepts
only keyword arguments in your defined method 'data_step'.
This method will process the inputs, and transform them into tensors.
Commonly used keyword arguments:
--------------------------------
graph: graphgallery graph classes.
graph_transform: string, Callable function,
or a tuple with function and dict arguments.
transform for the entire graph, it is used first.
device: device for preparing data, if None, it defaults to `self.device`
adj_transform: string, Callable function,
or a tuple with function and dict arguments.
transform for adjacency matrix.
attr_transform: string, Callable function,
or a tuple with function and dict arguments.
transform for attribute matrix.
other arguments (if have) will be passed into method 'data_step'.
"""
self.graph = gf.get(graph_transform)(graph)
cfg = self.cfg.data
if device is not None:
self.data_device = gf.device(device, self.backend)
else:
self.data_device = self.device
cfg.device = device
_, kwargs = gf.wrapper(self.data_step)(**kwargs)
kwargs['graph_transform'] = graph_transform
cfg.merge_from_dict(kwargs)
for k, v in kwargs.items():
if k.endswith("transform"):
setattr(self.transform, k, gf.get(v))
return self
def build(self, **kwargs):
"""This method is used for build your model, which
accepts only keyword arguments in your defined method 'model_step'.
Note:
-----
This method should be called after `process`.
Commonly used keyword arguments:
--------------------------------
hids: int or a list of them,
hidden units for each hidden layer.
acts: string or a list of them,
activation functions for each layer.
dropout: float scalar,
dropout used in the model.
lr: float scalar,
learning rate used for the model.
weight_decay: float scalar,
weight decay used for the model weights.
bias: bool,
whether to use bias in each layer.
use_tfn: bool,
this argument is only used for TensorFlow backend, if `True`, it will decorate
the model training and testing with `tf.function` (See `graphgallery.nn.modes.TFKeras`).
By default, it was `True`, which can accelerate the training and inference, by it may cause
several errors.
other arguments (if have) will be passed into your method 'model_step'.
"""
if self._graph is None:
raise RuntimeError("Please call 'trainer.make_data(graph)' first.")
use_tfn = kwargs.get("use_tfn", True)
if self.backend == "tensorflow":
with tf.device(self.device):
self.model, kwargs = gf.wrapper(self.model_step)(**kwargs)
if use_tfn:
self.model.use_tfn()
else:
kwargs.pop("use_tfn", None)
model, kwargs = gf.wrapper(self.model_step)(**kwargs)
self.model = model.to(self.device)
self.cfg.model.merge_from_dict(kwargs)
return self
def setup_graph(self, graph, graph_transform=None, device=None, **kwargs):
"""This method is used for process your inputs, which accepts
only keyword arguments in your defined method 'data_step'.
This method will process the inputs, and transform them into tensors.
Commonly used keyword arguments:
--------------------------------
graph: graphgallery graph instance.
the input graph
graph_transform: string, Callable function,
or a tuple with function and dict arguments.
transform for the entire graph, it is used first.
device: device for preparing data, if None, it defaults to `self.device`
adj_transform: string, Callable function,
or a tuple with function and dict arguments.
transform for adjacency matrix.
feat_transform: string, Callable function,
or a tuple with function and dict arguments.
transform for attribute (feature) matrix.
other arguments (if have) will be passed into method 'data_step'.
"""
self.cache_clear()
attr_transform = kwargs.pop("attr_transform", None)
if attr_transform:
warnings.warn("Argument 'attr_transform' is deprecated and will removed in future version, "
"please use 'feat_transform' instead.")
kwargs['feat_transform'] = attr_transform
self.graph = gf.get(graph_transform)(graph)
if device is not None:
self.data_device = torch.device(device)
else:
self.data_device = self.device
_, kwargs = gf.wrapper(self.data_step)(**kwargs)
kwargs['graph_transform'] = graph_transform
for k, v in kwargs.items():
if k.endswith("transform"):
setattr(self.transform, k, gf.get(v))
return self
def build(self, **kwargs):
"""This method is used for build your model, which
accepts only keyword arguments in your defined method 'model_step'.
Note:
-----
This method should be called after `setup_graph`.
Commonly used keyword arguments:
--------------------------------
hids: int or a list of them,
hidden units for each hidden layer.
acts: string or a list of them,
activation functions for each layer.
dropout: float scalar,
dropout used in the model.
lr: float scalar,
learning rate used for the model.
weight_decay: float scalar,
weight decay used for the model weights.
bias: bool,
whether to use bias in each layer.
other arguments (if have) will be passed into your method 'model_step'.
"""
if self._graph is None:
raise RuntimeError("Please call 'trainer.setup_graph(graph)' first.")
model, kwargs = gf.wrapper(self.model_step)(**kwargs)
self._model = model.to(self.device)
self.optimizer = self.config_optimizer()
self.scheduler = self.config_scheduler(self.optimizer)
self.loss = self.config_loss()
metrics = self.config_metrics()
if not isinstance(metrics, list):
metrics = [metrics]
self.metrics = metrics
return self
def build(self, **kwargs):
self.model, kwargs = gf.wrapper(self.model_builder)(**kwargs)
self.cfg.model.merge_from_dict(kwargs)
self.classifier = self.classifier_builder()
return self
