Apply the Sqrt operator entrywise.

Apply the Max operator entrywise.

Scale each input value by a constant value (c*x).

@class type_erased_matrix * @brief A type-erased wrapper around an @c El::Matrix<T,Device::CPU> * * @warning This class is an implementation deta

* Abstract base class for transforms on data. * * A transform takes a CPUMat and modifies it in-place. Transforms should * be thread-safe, as one i

Resize an image. */

Apply the Exp operator entrywise.

Apply the Mod operator entrywise.

Convolution -> Batch normalization -> ReLU Basic unit for ResNets. Assumes image data in NCHW format.

Convolution -> Batch normalization -> ReLU Basic unit for ResNets. Assumes image data in NCHW format.

Simple data structure to describe an LC system.

Basic block for channelwise fully-connected neural networks. Applies a dense linearity channelwise and a nonlinear activation function.

A simple size configuration entry for GPT-3 models.

See https://pytorch.org/docs/stable/generated/torch.nn.LayerNorm.html

Apply the Min operator entrywise.

Crop an image at the center. */

Crop an image at a random location. */

Downsample a 2D tensor using average pooling.

Shift zero-frequency component of discrete Fourier transform to center of spectrum. The input can have any dimension, although the first dime

Custom Dataset loader for MOF data.

Compute average pixel value w.r.t. distance from image center. We compute the distance between each image pixel and the image center. These d

Nearest-neighbor upsample a 2D tensor by a factor of 2 using the native gather layer.

Apply a random affine transform to an image. */

Apply the Cosh operator entrywise.

Building block for decoder in Transformer model. Comprised of two multi-head attention modules and a fully-connected feedforward network, eac

* Normalize with mean and standard deviation. * This is done channel-wise for images. If the input does not have channels, * (e.g. it is not an imag

* Extract a crop of random size and aspect ratio, then crop to a size. * This is commonly used for Inception-style networks and some other * image c

Resize an image then extract a random crop. */

Resize an image and then crop its center. */

Apply the Cos operator entrywise.

Basic block for fully-connected neural networks. Applies a dense linearity and a nonlinear activation function.

Apply the Pow operator entrywise.

Simple data structure to describe an OLCF system.

Apply the Tanh operator entrywise.

Building block for encoder in Transformer model. Comprised of multi-head attention and a fully-connected feedforward network, each with a res

Basic block of an optional concatenation layer and a list of 3D convolutional layers.

Trainable parameters for neural network.

Horizontally flip image data with given probability. */

Vertically flip image data with given probability. */

CosmoFlow neural network. See: Amrita Mathuriya, Deborah Bard, Peter Mendygral, Lawrence Meadows, James Arnemann, Lei Shao,

GCN Conv later. See: https://arxiv.org/abs/1609.02907

Graph Conv layer. See: https://arxiv.org/abs/1609.02907

LeNet-5, as presented in Y. LeCun et al., "Gradient-Based Learning Applied to Document Recognition", Proc. IEEE 1998.

Parallel instances of scaled dot-product attention. See: Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan

The strategy for mutation after a tournament in LTFB. When a trainer loses in a LTFB tournament, the winning model is copied ov

An opaque PyTorch module that prohibits Dynamo from inlining the contents of the original module, which might not be compilable.

Positional encoding, adapted from https://pytorch.org/tutorials/beginner/transformer_tutorial.html for batch_first operation.

Apply the Sin operator entrywise.

Apply the Sinh operator entrywise.

Building block for encoder in Transformer model. Comprised of multi-head attention and a fully-connected feedforward network, each with a res

@brief Class for LBANN batch inference algorithms. * * This execution algorithm is meant for running inference using a trained * model and sample

Convert an image from grayscale to color. */

Convert an image to grayscale. */

@brief Wrapper around a Python object pointer. * * Manages the reference count for a @c PyObject pointer and is * implicitly convertible to the p

@class parse_error * @brief std::exception subclass that is thrown if the parser * can not parse the arguments. */

Periodically print computational results. * Prints average objective function value and metric scores after * each training epoch and evaluation.

* Convert data to LBANN's native data layout. * Currently only supports converting from OpenCV layouts. * This will also rescale data from [0, 255]

Apply the Acos operator entrywise.

Apply the Acosh operator entrywise.

Apply the Asin operator entrywise.

Apply the Asinh operator entrywise.

Apply the Atan operator entrywise.

Apply the Atanh operator entrywise.

@brief Indicates that an invalid pattern is detected. */