brainpy.nn.runners.BPTT

class brainpy.nn.runners.BPTT(target, loss, optimizer=None, max_grad_norm=None, shuffle_data=True, jit=True, **kwargs)

The trainer implementing back propagation through time (BPTT) algorithm for recurrent neural networks.

__init__(target, loss, optimizer=None, max_grad_norm=None, shuffle_data=True, jit=True, **kwargs)

Methods

__init__(target, loss[, optimizer, ...])
f_grad([shared_kwargs])	Get gradient function.
f_loss([shared_kwargs])	Get loss function.
f_train([shared_kwargs])	Get training function.
fit(train_data[, test_data, num_batch, ...])	Fit the target model according to the given training and testing data.
predict(xs[, forced_states, ...])	Predict a series of input data with the given target model.

Attributes

mapping_type	Mapping type for the output and the target.
train_losses	Training loss.