def sequential_valid_loader(c: NLPAutoRegressionConfigs):
"""
### Sequential validation data loader
"""
return SequentialDataLoader(text=c.text.valid,
dataset=c.text,
batch_size=c.batch_size,
seq_len=c.seq_len)
def sequential_train_loader(c: NLPAutoRegressionConfigs):
"""
### Sequential training data loader
"""
return SequentialDataLoader(text=c.text.train,
dataset=c.text,
batch_size=c.batch_size,
seq_len=c.seq_len)
def valid_loader(c: Configs):
"""
Create a sequential data loader for validation
"""
return SequentialDataLoader(text=c.text.valid,
dataset=c.text,
batch_size=c.batch_size,
seq_len=c.seq_len)
def train_loader(c: Configs):
"""
Create a sequential data loader for training
"""
return SequentialDataLoader(text=c.text.train,
dataset=c.text,
batch_size=c.batch_size,
seq_len=c.seq_len)
def init(self):
# Create a configurable optimizer.
# Parameters like learning rate can be changed by passing a dictionary when starting the experiment.
optimizer = OptimizerConfigs()
optimizer.parameters = self.model.parameters()
optimizer.d_model = self.transformer.d_model
optimizer.optimizer = 'Noam'
self.optimizer = optimizer
# Create a sequential data loader for training
self.train_loader = SequentialDataLoader(text=self.text.train,
dataset=self.text,
batch_size=self.batch_size,
seq_len=self.seq_len)
# Create a sequential data loader for validation
self.valid_loader = SequentialDataLoader(text=self.text.valid,
dataset=self.text,
batch_size=self.batch_size,
seq_len=self.seq_len)
self.state_modules = [self.accuracy]
def train_loader(c: Configs):
return SequentialDataLoader(text=c.text.valid,
dataset=c.text,
batch_size=c.batch_size,
seq_len=c.seq_len)
