{"success":true,"data":{"author":"Thuerey group, Technical University of Munich","bucketAccessType":"public","bucketName":"thuerey-group/INC_Data","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-16T02:38:29.612Z","description":"Simulation data released by the Thuerey group (TU Munich) with INC, an indirect neural corrector for autoregressive hybrid PDE solvers (NeurIPS 2025), for training and evaluating learned corrections to coarse numerical solvers. Two systems are included. The Kuramoto-Sivashinsky equation (1D chaotic dynamics, periodic, 64 grid points, dt = 0.01) has 27 training, 3 validation and 6 test trajectories of 10,001 steps with varying domain sizes, stored as PyTorch .pth dictionaries. The backward-facing step (BFS) is 2D incompressible Navier-Stokes flow in a five-block channel geometry at Reynolds numbers 1300 to 1400 and step heights 0.85 to 1.0, downsampled to about 128 x 32 and saved every 0.1 time units as JSON metadata plus NPZ tensors per step (velocity, pressure, vertex coordinates, boundary conditions), with 3 training configurations of about 801 steps, 1 validation and a 3,001-step test run. A Results folder holds uncorrected baseline and INC-corrected rollouts for BFS. Layout is KS/Dataset/{train,valid,test} and BFS/Dataset/{train,valid,test}; about 3 GB. Use with the tum-pbs/INC code and PICT solver. Released under Apache-2.0 by the authors.","displayTitle":"INC Dataset: Kuramoto-Sivashinsky and Backward-Facing Step Flow Trajectories for Neural PDE Solver Correction","id":"6d821f4d-26f7-574e-bc90-06d789294375","license":"Apache-2.0","licenseEvidenceUrl":"https://huggingface.co/datasets/thuerey-group/INC_Data","objectCountEstimate":11449,"probeStatus":"ok","referenceUrl":"https://openreview.net/forum?id=s3Uk3lrfjy","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"thuerey-group/INC_Data\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"thuerey-group/INC_Data\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=thuerey-group, license:apache-2.0, size_categories:1K<n<10K, format:json, downloads=197"],"format":"mixed","generatedAt":"2026-11-16T02:38:34.954Z","layout":{"example":"BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.npz","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"504b030414000000080000002100fd9cd25a480000008400000005001400302e6e707901001000840000000000000048000000000000009bec17ea1b10c9c850c650ad9e925a9c5ca46ea5a06e9366a2aea3a09e965f54529498179f5f94920a12774bcc294e058a17672416a402f91a863a9a3a0ab50a1400aee776bbad00504b030414000000080000002100ad033f17e90700008008000005001400312e6e7079010010008008000000000000e90700000000000095d1fb3b55f91e07f0c82d4f6e9b08a7a1185214b9c75a1fcd8c7267bb4d24d7e87e43d749ecb051a7a422936453bbdca2dacb5e6b7dbfbb954b3974a60b8a622ba4db8866549aa929b3e7fc07e7f33cafe7f3bcdfcfe7b7cfc9a088c09028a519bb67fc64bd2e392d29d57ab9b9b5678a93b5adb975caf6d4f4d4846d71db53d725ffd3af4cd89296ace8d33624ec4856e4850eb6e6cb6ccd1d5cfe61636b9e61fe7f8fa6dd5239316a2327d242e5c43ab19c485a3e48541b3c21ba1d9f12cbc543c4db3523c4fe95a3c4a1c0174457e12b625c7b8cd83e7b8288899d245e847c26be995f436a2a94382976720d99c8d69022d75ab2e0692de92da923d7d5d69346e832693bd840d66a5f21a5fcab6468e33532642945da6d6f225ded79f069310f82dc79101cc603cf6c1e58dde1c103077de86ed4879b2b0c606cc80004f97320cad110b841433828","encoding":"hex","key":"BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.npz","truncatedAt":512}],"sampleObjects":["BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.json","BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.npz","BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3001.json","BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3001.npz","BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3002.json"],"sidecars":[{"bytes":"---\nlicense: apache-2.0\ntask_categories:\n  - time-series-forecasting\n  - tabular-regression\ntags:\n  - physics\n  - pde\n  - fluid-dynamics\n  - simulation\n  - numerical-methods\nsize_categories:\n  - 10K<n<100K\n---\n\n# INC Dataset: Implicit Neural Correction for PDE Solvers\n\n## Dataset Description\n\nThis dataset contains simulation data for training and evaluating implicit neural correction methods for partial differential equation (PDE) solvers. The dataset includes two challenging dynamical systems demonstrating complex spatiotemporal behaviors:\n\n1. **Kuramoto-Sivashinsky (KS) Equation** - 1D chaotic dynamics\n2. **Backward-Facing Step (BFS) Flow** - 2D incompressible Navier-Stokes with complex geometry\n\n### Dataset Summary\n\n- **Repository**: [INC: Implicit Neural Correction for PDE Solvers](https://github.com/tum-pbs/INC)\n- **Paper**: [INC: An Indirect Neural Corrector for Auto-Regressive Hybrid PDE Solvers](https://openreview.net/forum?id=s3Uk3lrfjy) (NeurIPS 2025)\n\n## Dataset Structure\n\n```\nINC_Data/\n├── KS/\n│   └── Dataset/\n│       ├── train/\n│       ├── valid/\n│       └── test/\n└── BFS/\n    └── Dataset/\n    │   ├── train/\n    │   ├── valid/\n    │   └── test/\n    └── Results/\n        └── NoModel/        # Baseline without correction\n        └── INC_SmallCNN/\n            └── {timestamp}_mstep8_.../ # an example model\n```\nThere are two main subdirectories corresponding to the two PDE systems, each containing training, validation, and test datasets. For BFS, there is also a `Results` directory showcasing baseline and corrected model results.\n\n### Data Fields\n\nEach dataset contains time-series simulation data with the following characteristics:\n\n\n#### Kuramoto-Sivashinsky Equation\n- **Spatial Resolution**: 64 grid points\n- **Temporal Resolution**: dt = 0.01\n- **Parameters**: Periodic boundary conditions\n- **File Format**: `.pth` (PyTorch dictionary with trajectories and metadata)\n- **Data Structure**: \n  - `trajectories`: shape `(num_trajectories, time_steps, spatial_points)` = `(27, 10001, 64)` for train\n  - `domain_size`: shape `(num_trajectories,)` = `(27,)` \n  - `metadata`: dict with generation parameters (gen_dt, resolution, time_scheme, etc.)\n- **Dataset Sizes**:\n  - Train: 27 trajectories × 10,001 timesteps\n  - Valid: 3 trajectories × 10,001 timesteps\n  - Test: 6 trajectories × 10,001 timesteps\n\n#### Backward-Facing Step (BFS)\n- **Spatial Resolution**: Multi-block grid with refinement, downsampled to approximately $128 \\times 32$\n- **Temporal Resolution**: Saved with fixed intervals (dt = 0.1)\n- **Physical Domain**: 2D channel with backward-facing step geometry (5 blocks)\n- **Parameters**: Reynolds numbers {1300, 1350, 1400}, Step height {0.85, 0.875, 1.0}\n- **File Format**: Combined `.json` (metadata) + `.npz` (tensor data) per timestep\n- **Data Structure**:\n  - Each configuration has 5 blocks with varying resolutions\n  - Block shapes vary by position: e.g., `(1, 2, 16, 16)` for velocity, `(1, 1, 16, 16)` for pressure\n  - Metadata includes: domain name, spatial dimensions, viscosity, block info, boundaries\n  - Data arrays: velocity (2 channels), pressure (1 channel), vertex coordinates, boundary conditions\n- **Dataset Sizes**:\n  - Train: 3 configurations × ~801 timesteps each\n  - Valid: 1 configuration × 301 timesteps\n  - Test: 1 configuration × 3,001 timesteps\n\n\n## Dataset Generation\n\nThe data was generated using classical numerical PDE solvers:\n\n- **Burgers**: 5th-order WENO scheme with RK4 time integration\n- **Kuramoto-Sivashinsky**: Pseudo-spectral method with exponential time differencing\n- **BFS**: PISO algorithm with custom CUDA implementation for multi-block domains\n\n### Generation Scripts\n\nThe original data generation code is available in the [INC repository](https://github.com/tum-pbs/INC):\n- `scripts/Sim_BFS.py` - Generate BFS flow data\n- `solvers/solver_1d.py` - Contains Burgers and KS solvers\n\n## Use Cases\n\nThis dataset is designed for:\n\n1. **Hybrid Physics-ML Models**: Training neural networks to correc","encoding":"utf8","key":"README.md","name":"README.md","size":6085,"truncated":true}],"sourceObjectKey":"BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.npz","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"thuerey-group/INC_Data\", filename=\"BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.npz\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"BFS/Dataset/test/Re1400_Base20-2_Step1.0_Dt0.1/DownBFSdomain_3000.npz","totalObjects":11449},"summary":"Simulation data released by the Thuerey group (TU Munich) with INC, an indirect neural corrector for autoregressive hybrid PDE solvers (NeurIPS 2025), for training and evaluating…","title":"INC Dataset: Kuramoto-Sivashinsky and Backward-Facing Step Flow Trajectories for Neural PDE Solver Correction","updatedAt":"2026-11-16T17:14:01.163Z","visibility":"public","websiteUrl":"https://github.com/tum-pbs/INC"}}