MexM-NN

Right now I am working with descriptors, to describe the environment of an atom using dscribe

Deep Learning in 3D

• https://towardsdatascience.com/intro-to-3d-deep-learning-e992f7efa6ee

• TensorFlow3D

• pytorch3d

  • https://pytorch3d.org/tutorials/

Installation of TensorFlow environment

Create a file requirements.txt with the following content:

tensorflow==2.5.0

tensorflow-datasets==4.3.0

Pillow==8.2.0

pandas==1.2.4

numpy==1.19.5

scipy==1.7.0

Create a virtual environment

virtualenv venv

In windows:

.\venv\Scripts\activate

In linux:

./venv/bin/activate

Install the requirements:

pip install -r requirements.txt

Refs:

• https://python.plainenglish.io/how-to-create-and-manage-virtual-environments-in-python-706bf22fc1b

Installation

https://docs.microsoft.com/en-us/windows/wsl/wsl2-install 

1. open a command prompt in elevated mode.  Link

   1. Click Start.

   2. In the search box, type cmd and then press Ctrl+Shift+Enter. If done properly, the below User Account Control window opens.

   3. Click Yes to run the Windows Command Prompt as Administrator.

2. install WSL2. Link

   1. dism.exe /online /enable-feature /featurename:Microsoft-Windows-Subsystem-Linux /all /norestart

   2. dism.exe /online /enable-feature /featurename:VirtualMachinePlatform /all /norestart

   3. install the distro you like from the microsoft store.

      1. Ubuntu-1804-lts 

      2. in Windows cmd line: wsl --setdefault Ubuntu-18.04 2

      3. Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Windows-Subsystem-Linux

[1]Microsoft Corporation. "Windows Subsystem for Linux Installation Guide for Windows 10" Link

What is MexM-NN?

• • MexM-NN is neural network package for build on top of MexM

  • Build on dscribe, tensorflow2, and scipy

  • https://pytorch.org/

    • The installation of pytorch is straightforward.

      • https://pytorch.org/get-started/locally/

      • conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

      • MacOS doesn't install CUDA

  • https://singroup.github.io/dscribe/