lbann
lbann copied to clipboard
LLNL
Livermore Big Artificial Neural Network Toolkit
This will hold us over for a transition period. I'd like to get to Catch2 v3 in general, though, as it's better factored. (I can do that now, but I...
- Enables distconv implementations of Scatter and Gather layers - Implements NVSHMEM based RMA kernels for scatter/gather on DiHydrogen tensors - Adds example applications in `applications/graph/DistConvGNN/synthetic` for benchmarking distributed Scatter,...
There is a poor behavior in Spack environments where PyTorch is concretizing some of its dependents with different GPU architectures than what LBANN and the rest of the environment are...
The cutout layer applies a fixed-size zero mask to a random location of each input image. The cutout layer selects a pixel coordinate within the image as a center point...
The LBANN Distconv adapter for layers mandates that only the first input tensor to distconv-enabled layer can be a non-DiHydrogen tensor. We raise an error if a tensor requires a...
Enables splitting along the 1st dimension in a 4D tensor and using distconv to perform StridedBatchedGEMMs. - Need to resolve issue #2126 prior to merging
Adding padding for vision in the transformation pipeline of the data reader. Padding the given image on all sides with 0 (black) value.
The `Identity` layer with a parallel strategy defined crashes on the head of [develop](https://github.com/LLNL/lbann/commit/c4152b4bedf5f590b7c9c61014c307dc1b9cd12b) on Lassen with distconv enabled. LBANN is built using the heads of Hydrogen, DiHydrogen, and Aluminium...
Lbann implementation for Roberta with masked language modeling using SMILE strings.
