bootstrap.Model

class bootstrap.models.model.Model(engine=None, cuda_tf=<class 'bootstrap.datasets.transforms.ToCuda'>, detach_tf=<class 'bootstrap.datasets.transforms.ToDetach'>, network=None, criterions={}, metrics={})

Model contains a network, two criterions (train, eval) and two metrics.

cpu()

Moves all model parameters and buffers to the CPU.

cuda(device_id=None)

Moves all model parameters and buffers to the GPU.

Parameters:device_id (int, optional) – if specified, all parameters will be copied to that device

eval()

Activate evaluation mode

forward(batch)

Prepare the batch and feed it to the network, criterion and metric.

Returns:a dictionary of outputs Return type:out (dict)

load_state_dict(state, *args, **kwgs)

prepare_batch(batch)

Prepare a batch with two functions: cuda_tf and detach_tf (only in eval mode)

state_dict(*args, **kwgs)

train()

Activate training mode

class bootstrap.models.model.DefaultModel(engine=None, cuda_tf=<class 'bootstrap.datasets.transforms.ToCuda'>, detach_tf=<class 'bootstrap.datasets.transforms.ToDetach'>)

An extension of Model that relies on factory calls

class bootstrap.models.model.SimpleModel(engine=None, cuda_tf=<class 'bootstrap.datasets.transforms.ToCuda'>, detach_tf=<class 'bootstrap.datasets.transforms.ToDetach'>)

An extension of DefaultModel that modifies the forward function

forward(batch)

The forward call to the network uses batch[‘data’] instead of batch