UCL-CCS / DRTNN

Deep Residual Transformer Neural Network (DRTNN)

[Deep Residual Transformer Neural Network (DRTNN)]

The Transformer Neural Network in this project is not depend on the data representation and the application (the surrogate model works for any type of linear and nonlinear systems). It can be easily modified for end user purposes.

The following image shows a brief comparison between the architectures of the Transformer Network in the current study and a typical Transformer Network.

Neural transformer network for predicting stress!

Transformer network, in the current form, is a computational learning system for predicting stress. It is a surrogate model for the computational model of a heterogeneous multi-scale atomistic continuum coupling system, SCEMa (https://GitHub.com/UCL-CCS/SCEMa).

Running the Program!

User needs first to install Anaconda https://www.anaconda.com/

Then

  - conda env create -f environment.yml
  or
  - conda create --name traintest --file Building_identical_conda_environment-file.txt

and

  - conda activate traintest

finally

  - python  main.py