nntoolbox.utils.utils module

nntoolbox.utils.utils.compute_num_batch(data_size: int, batch_size: int)

Compute number of batches per epoch

Parameters

• data_size – number of datapoints

• batch_size – number of datapoints per batch

Returns

nntoolbox.utils.utils.copy_model(model: torch.nn.modules.module.Module) → torch.nn.modules.module.Module

Return an exact copy of the model (both architecture and initial weights, without tying the weights)

Parameters

model – model to be copied

Returns

a copy of the model

nntoolbox.utils.utils.count_trainable_parameters(model: torch.nn.modules.module.Module) → int

Based on https://discuss.pytorch.org/t/how-do-i-check-the-number-of-parameters-of-a-model/4325/8

Parameters

model –

Returns

nntoolbox.utils.utils.dropout_mask(t: torch.Tensor, size: Tuple[int, …], drop_p)

nntoolbox.utils.utils.find_index(array, value)

nntoolbox.utils.utils.get_all_submodules(module: torch.nn.modules.module.Module) → List[torch.nn.modules.module.Module]

Get all submodules of a module

Parameters

model –

Returns

list of all submodules of a model

nntoolbox.utils.utils.get_children(model: torch.nn.modules.module.Module) → List[torch.nn.modules.module.Module]

Parameters

model –

Returns

list of all children of a model

nntoolbox.utils.utils.get_device()

Convenient helper for getting device

Returns

a torch device object (gpu if exists)

nntoolbox.utils.utils.get_trainable_parameters(model: torch.nn.modules.module.Module) → List[torch.Tensor]

nntoolbox.utils.utils.is_nan(tensor: torch.Tensor) → bool

Check if any element of a tensor is NaN

Parameters

tensor –

Returns

whether any element of the tensor is NaN

nntoolbox.utils.utils.is_valid(tensor: torch.Tensor) → bool

Check if a tensor is valid (not inf + not nan)

Parameters

tensor –

Returns

whether a tensor is valid

nntoolbox.utils.utils.load_model(model: torch.nn.modules.module.Module, path: str)

Load the model from path

:param model :param path: path of saved model

nntoolbox.utils.utils.save_model(model: torch.nn.modules.module.Module, path: str)

Save a model

Parameters

• model –

• path – path to save model at

nntoolbox.utils.utils.to_onehot(label: torch.Tensor, n_class: Optional[int] = None) → torch.Tensor

Return one hot encoding of label (assuming the label index is 1)

Parameters

• label –

• n_class –

Returns

nntoolbox.utils.utils.to_onehotv2(label: torch.Tensor, n_class: Optional[int] = None) → torch.Tensor

Return one hot encoding of label (assuming the label index is -1)

Parameters

• label –

• n_class –

Returns