@details Forward prop is applied to all layers, except input * layers. It is assumed that input layers have already loaded data. */

Create a new matrix. */

Return basename without extention.

Return file extention name.

(trb 10/03/2023): used

@brief Increment the current step in the training algorithm * @details Increment the step count in the training * algorithm's internal state

Marks the given tensors as input of the tested LBANN model, and returns a list of matching LBANN Slice nodes, potentially reshaped to

----------------------------------------------- Weight sharding functions ----------------------------------------------- Are weights sharded across r

only used in LTFB; from that file: "Note that this is a temporary fix for the current use of the tournament"

only used in LTFB; from that file: "Note that this is a temporary fix for the current use of the tournament"

Returns the number of elements in a NumPy array, PyTorch array, or integer.

Base path to parallel file system.

@brief Set the model that manages this layer. */

Submit batch job. The script file is written before being submitted. Args: overwrite (bool): Whether to overwrite script

@brief Construct with weights names * @param[in] weights_names Names of weights to exchange. If * empty, then all

Places the features according the indices to an expanded matrix output[i] = features[indices[i]] Args: features (Layer) :

Performs a sum-reduction of the features according the indices. output[indices[i]] += features[i] Args: features (layer) :

Initialize convolution module. Args: num_dims (int): Number of dimensions. out_channels (int): Number of output chann

Xavier uniform initializer Args: fan_in (int): input size of the learning layer fan_out (int): output size of the learning layer Returns: (U

@todo Deprecated.