Model

class marius.nn.Model

__init__(self: marius._nn.Model, arg0: GeneralEncoder, arg1: Decoder, arg2: marius._nn.LossFunction, arg3: Reporter) → None

__init__(self: marius._nn.Model, encoder: GeneralEncoder, decoder: Decoder, loss: marius._nn.LossFunction = None, reporter: Reporter = None, sparse_lr: float = 0.1) → None

broadcast(self: marius._nn.Model, devices: List[torch.device]) → None

forward_lp(self: marius._nn.Model, batch: marius._data.Batch, train: bool) → Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]

forward_nc(self: marius._nn.Model, node_embeddings: Optional[torch.Tensor], node_features: Optional[torch.Tensor], dense_graph: marius._data.DENSEGraph, train: bool) → torch.Tensor

all_reduce(self: marius._nn.Model) → None

clear_grad(self: marius._nn.Model) → None

clear_grad_all(self: marius._nn.Model) → None

property decoder

property device

property device_models

property encoder

evaluate_batch(self: marius._nn.Model, batch: marius._data.Batch) → None

property learning_task

load(self: marius._nn.Model, directory: str, train: bool) → None

property loss_function

property optimizers

property reporter

save(self: marius._nn.Model, directory: str) → None

property sparse_lr

step(self: marius._nn.Model) → None

step_all(self: marius._nn.Model) → None

train_batch(self: marius._nn.Model, batch: marius._data.Batch, call_step: bool = True) → None