DAML

Data Assimilation and Machine Learning integration framework

Dependencies

This project assumes that the below dependencies have already been built.

spack-stack 1.5.1

On HPC, just load the GDASApp modules

JEDI

Just build your favorite jedi-bundle. This project makes use of oops, ufo, ioda, vader, saber, soca, atlas, eckit, ...

Torch/PyTorch

Cloning and building pytorch

git clone --recursive --branch v2.3.1 https://github.com/pytorch/pytorch
cd pytorch
mkdir build
cd build
cmake -DUSE_CUDA=OFF -DUSE_CUDNN=OFF -DUSE_NCCL=OFF -DBUILD_CAFFE2_OPS=OFF -DUSE_MKLDNN=OFF -DUSE_DISTRIBUTED=ON -DCMAKE_INSTALL_PREFIX=<path to install> ..
make -j<n>
make install

Applications

Bias/Error correction for MV based sea surface salinity retrievals

Coming soon ...

Unbalanced background error for sea surface height

Estimate the unblanced part of the background error for sea surface height. Naive implementation that demonstrates the use of the JEDI and Torch libraries.

Build process

For now, you will have to specify the root instalation of the JEDI repositories and Torch:

mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=../install \
      -DTorch_ROOT=<path to pytorch install> \
      -Doops_ROOT=<path to oops install> \
      -Datlas_ROOT=<path to atlas install> \
      ..
make -j<n>
make install