Skip to content

suyashbire1/oceanfourcast

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 

Repository files navigation

OceanFourcast

Can transformer methods be used to create fast emulators for forward and partial derivative computations in ocean modeling?

Installation

git clone [email protected]:suyashbire1/oceanfourcast.git
cd oceanfourcast
conda create --name oceanfourcast
conda activate
pip install -e .

Download data

mkdir -p data/processed/
scp [email protected]:/path/to/file/mitgcm/double_gyre/run3/dynDiag_subset.nc data/processed/. # Sample dataset
scp [email protected]:/path/to/file/mitgcm/double_gyre/run3/dynDiag.nc data/processed/. # Full dataset
python oceanfourcast/load_numpy.py --xarray_data_file "data/processed/unet/dynDiag_subset.nc" # Convert .nc to .npy

Train

python oceanfourcast/train.py --data_file "data/processed/dynDiags.npy" --batch_size 2

Train baseline models

# UNet
python oceanfourcast/train_unet.py --modelstr "unet" --data_file "data/processed/unet/dynDiags.npy" --batch_size 2 --output_dir "models/temp/mitgcm/unet/"