Poseidon
Collection
The Poseidon family of PDE foundation models + their pretraining data (from PDEGym). • 10 items • Updated • 5
Quick Links
The Large model of the Poseidon family with approx. 629M parameters, see https://arxiv.org/abs/2405.19101.
It is trained on a computational grid of size 128 x 128 and expects density, velocity in horizontal direction, velocity in vertical direction, and pressure as inputs.
This is a research model.
- Downloads last month
- 400
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
# Load model directly from transformers import AutoImageProcessor, SwinForPDE processor = AutoImageProcessor.from_pretrained("camlab-ethz/Poseidon-L") model = SwinForPDE.from_pretrained("camlab-ethz/Poseidon-L", device_map="auto")