Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -19,8 +19,9 @@ tags:
|
|
| 19 |
|
| 20 |
WaveFlow is a conditional flow-matching model for ensemble prediction of
|
| 21 |
significant wave height, peak period, mean direction and water level on
|
| 22 |
-
unstructured coastal/global meshes, conditioned on atmospheric fields
|
| 23 |
-
|
|
|
|
| 24 |
|
| 25 |
This repository hosts the three final checkpoints, the mesh/statistics assets
|
| 26 |
and the figure source data. The model, training loop, sampling code and exact
|
|
@@ -55,8 +56,9 @@ python reproduce.py infer fig3f # example: global Hs, 10 members
|
|
| 55 |
```
|
| 56 |
|
| 57 |
For your own inputs, run `main.py --mode sample` with the released config and
|
| 58 |
-
checkpoint;
|
| 59 |
-
|
|
|
|
| 60 |
documented in `TRAINING.md`.
|
| 61 |
|
| 62 |
## Files and checksums
|
|
|
|
| 19 |
|
| 20 |
WaveFlow is a conditional flow-matching model for ensemble prediction of
|
| 21 |
significant wave height, peak period, mean direction and water level on
|
| 22 |
+
unstructured coastal/global meshes, conditioned on atmospheric fields (ERA5
|
| 23 |
+
reanalysis, IFS ensemble forecasts, or AI weather forecasts such as GenCast)
|
| 24 |
+
and optional tide phase.
|
| 25 |
|
| 26 |
This repository hosts the three final checkpoints, the mesh/statistics assets
|
| 27 |
and the figure source data. The model, training loop, sampling code and exact
|
|
|
|
| 56 |
```
|
| 57 |
|
| 58 |
For your own inputs, run `main.py --mode sample` with the released config and
|
| 59 |
+
checkpoint; `-o sampling.atm_source=gencast` switches to AI forecast forcing
|
| 60 |
+
(see `TRAINING.md` for the per-member file layout). See the GitHub `README.md`
|
| 61 |
+
and `WEIGHTS.md` for the exact command, data layout and GPU requirements. Training/fine-tuning on your own hindcast is
|
| 62 |
documented in `TRAINING.md`.
|
| 63 |
|
| 64 |
## Files and checksums
|