Cloudflow

Warning

Experimental. This repository is under active development and has not stabilised; expect breaking changes between commits. Raise an issue if you encounter problems.

CloudFlow is a conditional flow-matching model that generates MODIS-like cloud radiances conditioned on ERA5 atmospheric states.

Install

Make sure you have uv installed on your system then run:

uv sync

Quickstart

from cloudflow import Conditioning, load_model, sample

model, info = load_model("hf://treigerm/cloudflow", device="cuda")
# Or load from a local directory:
# model, info = load_model("checkpoints/cloudflow", device="cuda")

c = info.crop_size  # e.g. 256
cond = Conditioning.from_modis_hdf(
    hdf_path="data/MYD021KM.A2023157.0820.061.hdf",
    data_info=info,
    bounds=(500, 500 + c, 800, 800 + c),
)

samples = sample(model, cond, num_ensembles=10, num_steps=30)
# samples is a torch.Tensor of shape (10, C_out, H, W).

Or, via the CLI:

uv run cloudflow sample \
    --ckpt checkpoints/cloudflow \
    --modis-hdf data/MYD021KM.A2023157.0820.061.hdf \
    --bounds 500,756,800,1056 \
    --out samples.nc