Tensors to device (!13) · Merge requests · Accelerated_training / accelerated_deeplearning_training · GitLab

During the maintenance period on Monday, 14 Oct 2024, all new projects will no longer create a container registry as part of the project area. Already existing projects will be unaffected at this time. Instead, users should utilize the official ORNL container registries at camden.ornl.gov (internal/moderate) and savannah.ornl.gov (external/low). Please see ORNL Today article for more info.

Lupo Pasini, Massimiliano requested to merge tensors_to_device into master Feb 03, 2021

tensor.to(device) used in "modules/NN_models.py" and "modules/optimizers.py" to guarantee that GPUs are used if present
number of runs introduced as parameter in input file. Each run fixed the random seed to the number of the iteration index
nonlinear synthetic dataset added in the "utils/dataloader.py"
LeakyReLU added as optional activation function to choose

Edited Feb 03, 2021 by Lupo Pasini, Massimiliano