| --- |
| license: apache-2.0 |
| language: |
| - en |
| tags: |
| - OneScience |
| - fluid-dynamics |
| library_name: pytorch |
| --- |
| |
| # RONOM |
|
|
| ## Model and experiment |
|
|
| This package reproduces the one-dimensional viscous Burgers input generalization experiment from the [paper](https://arxiv.org/abs/2507.12814). RONOM combines a convolutional reduced-order encoder, a latent Neural ODE, and a decoder. This experiment uses 33 spatial points, 11 time points, four latent channels, and fixed-step RK4 integration. |
|
|
| The package includes model code, `config/config.yaml`, training, inference and visualization scripts, the recorded `results/`, and the selected `weight/best_model.pth` checkpoint. |
|
|
| ## Data |
|
|
| Download [OneScience/RONOM-Burgers](https://www.modelscope.cn/datasets/OneScience/RONOM-Burgers) with `modelscope download --dataset OneScience/RONOM-Burgers --local_dir ./data`. Burgers.mat contains 1,000 trajectories: input [1000,1025], output [1000,101,1025], tspan [1,101], float64. Split: 600 train, 200 validation, 200 test. |
|
|
| Original dataset: https://zenodo.org/records/21451819 Download the dataset before running inference where an external dataset is required. The release configuration uses relative paths, not SCNet personal-directory paths. |
|
|
| ## Installation and usage |
|
|
| ```bash |
| hf download OneScience-Group/RONOM --local-dir ./RONOM |
| cd RONOM |
| pip install torch numpy scipy h5py pyyaml matplotlib |
| python scripts/inference.py |
| python scripts/result.py |
| ``` |
|
|
| For a fresh training run, use an allocated GPU/DCU compute node with a compatible PyTorch build, keep a separate copy of the supplied checkpoint, then run `python scripts/train.py` from the package root. The full budget is specified in `config/config.yaml`. Cluster-specific Slurm scripts, if present, need adaptation outside the original cluster. |
|
|
| ## Training and reported result |
|
|
| The run completed the configured 5,000 epochs. The first 600 trajectories were used for training, the next 200 for validation, and the final 200 for an independent test. |
|
|
| Mean test-trajectory RMSE was 0.00170540 versus 0.00134 in paper Table 1(a), about 27.3% higher. Mean relative L2 was 0.00986709. See `results/` for the recorded metrics and visualizations. The supplied checkpoint comes from a completed run; another run may differ. |
|
|
| ## Source and license |
|
|
| - Paper: https://arxiv.org/abs/2507.12814. |
| - Dataset source: https://zenodo.org/records/21451819. |
| - This model package uses the license indicated in the card metadata. Dataset redistribution terms, where not separately stated, must not be inferred from the code license. |
|
|
