{"success":true,"data":{"author":"Neil Ashton, Danielle Maddix, Samuel Gundry, Parisa Shabestari (caemldatasets.org)","bucketAccessType":"public","bucketName":"neashton/ahmedml","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-14T14:15:22.986Z","description":"AhmedML contains 500 geometric variations of the Ahmed car body, a simplified car-like shape that reproduces flow features of road vehicles such as geometry- and pressure-induced separation, recirculation and vortex shedding. Each case was run transiently for about 80 convective time units with hybrid RANS-LES in OpenFOAM v2212 on meshes of about 20 million cells, by Neil Ashton, Danielle Maddix, Samuel Gundry and Parisa Shabestari. Measured 2026-11-14, the repository root holds one run_N folder per geometry; each contains the STL geometry, geometry parameters (CSV), a boundary VTP surface file with native polygon areas (NPY), a volume VTU of about 5 GB, x/y/z slices, time-averaged drag and lift coefficients and images, and the root also has all-run CSV tables and the SolveSpace parametric geometry file. The publisher lists about 3.1 TB stored. It is used to benchmark machine learning surrogates for surface pressure, wall shear stress and integrated forces. Open VTP/VTU with PyVista, VTK or ParaView. Licensed CC-BY-SA-4.0; cite Ashton et al., arXiv:2407.20801.","displayTitle":"AhmedML: High-Fidelity CFD Dataset for Low-Speed Bluff-Body Aerodynamics","id":"7d288aad-76e3-5f7d-936b-c7b4613fa044","license":"CC-BY-SA-4.0","licenseEvidenceUrl":"https://caemldatasets.org/ahmedml/","objectCountEstimate":61528,"probeStatus":"ok","referenceUrl":"https://arxiv.org/abs/2407.20801","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"neashton/ahmedml\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"neashton/ahmedml\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=neashton, license:cc-by-sa-4.0, downloads=18361"],"format":"mixed","generatedAt":"2026-11-14T14:15:36.640Z","layout":{"example":"force_mom_all.csv","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"run, cd, cl\n1,2.3848566658e-01, -9.4516081781e-02\n2,3.0541930003e-01, 1.5083035464e-01\n3,2.4047351597e-01, 2.2527410672e-01\n4,3.8193639963e-01, 1.1461028679e-01\n5,2.5481603392e-01, -3.3455548361e-02\n6,3.6013296383e-01, 4.8948631673e-01\n7,3.3543675429e-01, -5.8771538003e-02\n8,3.7182129039e-01, 3.8785918709e-01\n9,3.5390064921e-01, -1.6458953732e-01\n10,1.8888523355e-01, 1.4873152968e-01\n11,3.2780176114e-01, -1.1852656482e-01\n12,3.8282940310e-01, 3.9790251136e-01\n13,3.5914224410e-01, 4.1769612416e-01\n14,3.52831","encoding":"utf8","key":"force_mom_all.csv","truncatedAt":512}],"sampleObjects":["ahmedml.slvs","force_mom_all.csv","force_mom_varref_all.csv","geo_parameters_all.csv","openfoam-casesetup.tgz"],"sidecars":[{"bytes":"Attribution-ShareAlike 4.0 International\n\n=======================================================================\n\nCreative Commons Corporation (\"Creative Commons\") is not a law firm and\ndoes not provide legal services or legal advice. Distribution of\nCreative Commons public licenses does not create a lawyer-client or\nother relationship. Creative Commons makes its licenses and related\ninformation available on an \"as-is\" basis. Creative Commons gives no\nwarranties regarding its licenses, any material licensed under their\nterms and conditions, or any related information. Creative Commons\ndisclaims all liability for damages resulting from their use to the\nfullest extent possible.\n\nUsing Creative Commons Public Licenses\n\nCreative Commons public licenses provide a standard set of terms and\nconditions that creators and other rights holders may use to share\noriginal works of authorship and other material subject to copyright\nand certain other rights specified in the public license below. The\nfollowing considerations are for informational purposes only, are not\nexhaustive, and do not form part of our licenses.\n\n     Considerations for licensors: Our public licenses are\n     intended for use by those authorized to give the public\n     permission to use material in ways otherwise restricted by\n     copyright and certain other rights. Our licenses are\n     irrevocable. Licensors should read and understand the terms\n     and conditions of the license they choose before applying it.\n     Licensors should also secure all rights necessary before\n     applying our licenses so that the public can reuse the\n     material as expected. Licensors should clearly mark any\n     material not subject to the license. This includes other CC-\n     licensed material, or material used under an exception or\n     limitation to copyright. More considerations for licensors:\n    wiki.creativecommons.org/Considerations_for_licensors\n\n     Considerations for the public: By using one of our public\n     licenses, a licensor grants the public permission to use the\n     licensed material under specified terms and conditions. If\n     the licensor's permission is not necessary for any reason--for\n     example, because of any applicable exception or limitation to\n     copyright--then that use is not regulated by the license. Our\n     licenses grant only permissions under copyright and certain\n     other rights that a licensor has authority to grant. Use of\n     the licensed material may still be restricted for other\n     reasons, including because others have copyright or other\n     rights in the material. A licensor may make special requests,\n     such as asking that all changes be marked or described.\n     Although not required by our licenses, you are encouraged to\n     respect those requests where reasonable. More considerations\n     for the public:\n    wiki.creativecommons.org/Considerations_for_licensees\n\n=======================================================================\n\nCreative Commons Attribution-ShareAlike 4.0 International Public\nLicense\n\nBy exercising the Licensed Rights (defined below), You accept and agree\nto be bound by the terms and conditions of this Creative Commons\nAttribution-ShareAlike 4.0 International Public License (\"Public\nLicense\"). To the extent this Public License may be interpreted as a\ncontract, You are granted the Licensed Rights in consideration of Your\nacceptance of these terms and conditions, and the Licensor grants You\nsuch rights in consideration of benefits the Licensor receives from\nmaking the Licensed Material available under these terms and\nconditions.\n\n\nSection 1 -- Definitions.\n\n  a. Adapted Material means material subject to Copyright and Similar\n     Rights that is derived from or based upon the Licensed Material\n     and in which the Licensed Material is translated, altered,\n     arranged, transformed, or otherwise modified in a manner requiring\n     permission under the Copyright and Similar Rights held by the\n     Licensor. For purposes of this Public License, where the Licensed\n     Material is a musical work, p","encoding":"utf8","key":"LICENSE.txt","name":"LICENSE.txt","size":20138,"truncated":true},{"bytes":"---\nlicense: cc-by-sa-4.0\nviewer: false \n---\nAhmedML: High-Fidelity Computational Fluid Dynamics dataset for incompressible, low-speed bluff body aerodynamics\n-------\n\nContact: \n----------\nNeil Ashton (NVIDIA) - contact@caemldatasets.org\n\nwebsite:\n----------\nhttps://caemldatasets.org\n\nSummary:\n-------\n\nThis dataset contains 500 different geometric variations of the Ahmed Car Body - a simplified car-like shape that exhibits many of the flow topologies that are present on bluff bodies such as road vehicles. The dataset contains a wide range of geometries that exhibit fundamental flow physics such as geometry and pressure-induced flow separation of flows as well as 3D vortical structures. Each variation of the Ahmed car body were run using a time-accurate hybrid Reynolds-Averaged Navier-Stokes (RANS) - Large-Eddy Simulation (LES) turbulence modelling approach using the open-source CFD code OpenFOAM. The dataset contains both surface boundary, 3D volume, geometry STL and forces/moments in open-source formats (.vtu,.vtp). \n\nCFD Solver:\n----------\nAll cases were run using the open-source finite-volume code OpenFOAM v2212. Each case was run transiently for approximately 80 convective time units (CTU) on meshes of approximately 20M cells.  Please see the paper for full details on the code and validation:\n\nHow to cite this dataset:\n----------------\nIn order to cite the use of this dataset please cite the paper below which contains full details on the dataset. It can be found here: https://arxiv.org/abs/2407.20801\n\n@article{ashton2024ahmed,\n    title = {{AhmedML: High-Fidelity Computational Fluid Dynamics dataset for incompressible, low-speed bluff body aerodynamics}},\n    year = {2024},\n    journal = {arxiv.org},\n    author = {Ashton, Neil and Maddix, Danielle and Gundry, Samuel and Shabestari, Parisa}\n}\n\nFiles:\n-------\nEach folder (e.g run_1,run_2...run_\"i\" etc) corresponds to a different geometry that contains the following files where \"i\" is the run number: \n* ahmed_i.stl : geometry stl (~5mb):\n* geo_parameters_1.csv (missing run 500): parameters that define the geometry\n* boundary_i.vtp : Boundary VTP (~500mb)\n* `boundary_cell_area_i.npy`: Native polygon areas (`m^2`) for `boundary_i.vtp`, in exact `CellData` tuple order.\n* volume_i.vtu : Volume field VTU (~5GB)\n* force_mom_i.csv : forces (Cd,Cl) time-averaged with constant reference area\n* force_mom_varref_i.csv : forces (Cd,Cl) time-averaged with varying reference area\n* slices : folder containing .vtp slices in x,y,z that contain flow-field variables\n* images : (folder) that contains images of the following variables (CpT, UxMean) for slices of the domain in the X,Y & Z locations.\nIn addition we provide:\n* force_mom_all.csv : run, cd,cl for all runs in a single file\n* force_mom_varref_all.csv : run, cd,cl for all runs in a single file with varying reference area\n* geo_parameters_all.csv : all the geometry parameters for each run inside a single file\n* ahmedml.slvs : SolveSpace input file to create the parametric geometries\n* stl : folder containing stl files that were used as inputs to the OpenFOAM process\n* openfoam-casesetup.tgz : complete OpenFOAM setup that can be used to extend or reproduce the dataset\n* validation : folder containing full outputs from all four mesh levels that were used to validate the methodology\n* [`splits/`](splits/): deterministic benchmark manifests, methods documentation, derived metrics, diagnostic figures, and generation code\n\n## Recommended dataset splits\n\nAhmedML provides eight deterministic train/validation/test split families in\n[`splits/manifest.json`](splits/manifest.json). Identifiers correspond to the\ntop-level `run_N` directories.\n\n| Split | Type | Train | Validation | Test | Intended evaluation |\n|---|---:|---:|---:|---:|---|\n| `full` | In-distribution | 400 | 50 | 50 | Seed-42 public baseline |\n| `medium` | In-distribution | 133 | 50 | 50 | Intermediate data efficiency |\n| `scarce` | In-distribution | 67 | 50 | 50 | Low-data evaluation |\n| `super_scarce` | In-distribution | 11 | 50 | 50 | Extreme low-data evaluation |\n| `geometry` ","encoding":"utf8","key":"README.md","name":"README.md","size":8191,"truncated":true},{"bytes":"# AhmedML dataset splits\n\nThis directory provides deterministic train/validation/test assignments for the\n[AhmedML](https://huggingface.co/datasets/neashton/ahmedml) dataset. The\nauthoritative assignments are stored in [`manifest.json`](manifest.json) as a\nflat JSON object. Keys follow the pattern `{split_name}_{train,val,test}`, and\neach value is a numerically sorted list of identifiers matching the top-level\n`run_N` directories.\n\nAhmedML contains 500 Ahmed-body variants, indexed from `run_1` to `run_500`.\nThe aggregate force and moment table contains all 500 runs. The aggregate\ngeometry-parameter table contains 499 rows because `run_500` is absent; the\ntreatment of this missing row is described below.\n\n## Splits at a glance\n\n| Split | Type | Train | Validation | Test | Intended evaluation |\n|---|---:|---:|---:|---:|---|\n| `full` | In-distribution | 400 | 50 | 50 | Seed-42 public baseline |\n| `medium` | In-distribution | 133 | 50 | 50 | Intermediate data efficiency |\n| `scarce` | In-distribution | 67 | 50 | 50 | Low-data evaluation |\n| `super_scarce` | In-distribution | 11 | 50 | 50 | Extreme low-data evaluation |\n| `geometry` | OOD | 350 | 50 | 100 | STL-surface geometry extrapolation |\n| `high_drag` | OOD | 350 | 50 | 100 | High-drag extrapolation |\n| `low_drag` | OOD | 350 | 50 | 100 | Low-drag extrapolation |\n| `image_wake` | OOD | 350 | 50 | 100 | Image-derived wake extrapolation |\n\nThe data-efficiency training sets form a strict nested sequence:\n\n`super_scarce_train ⊂ scarce_train ⊂ medium_train ⊂ full_train`\n\nThey use the same validation and test assignments. For every\nout-of-distribution (OOD) family, validation is sampled from the training-side\npopulation; the held-out extreme is reserved for final testing.\n\n## Using the committed manifest\n\nNormal benchmark use requires only the committed manifest. Regenerating the\nsplits is not required.\n\n```python\nimport json\nfrom pathlib import Path\n\nmanifest = json.loads(Path(\"splits/manifest.json\").read_text())\n\ntrain_ids = manifest[\"geometry_train\"]\nval_ids = manifest[\"geometry_val\"]\ntest_ids = manifest[\"geometry_test\"]\n```\n\nChange the `geometry` prefix to `full`, `medium`, `scarce`,\n`super_scarce`, `high_drag`, `low_drag`, or `image_wake` to select another\nfamily.\n\nDownload only the split package with:\n\n```bash\nhf download neashton/ahmedml \\\n  --type dataset \\\n  --include \"splits/**\" \\\n  --local-dir ./ahmedml\n```\n\nValidation data may be used for model and hyperparameter selection. Test data\nshould be reserved for final evaluation and should not inform normalization,\nfeature design, or repeated visual inspection during development.\n\n## Construction principles\n\n1. **Stable public baseline.** The `full` family preserves the established\n   seed-42 AhmedML assignment used by\n   [Noether](https://github.com/Emmi-AI/noether/blob/main/src/noether/data/datasets/cfd/caeml/ahmedml/split.py).\n2. **In-distribution validation.** OOD validation cases are selected from the\n   training-side population rather than the extreme test region.\n3. **Nested data-efficiency subsets.** Smaller training sets are strict\n   subsets of larger sets, with validation and test held fixed.\n4. **Direct geometry comparison.** The geometry OOD score is computed from STL\n   surfaces rather than inferred only from geometry parameters.\n5. **Dataset-defined physical quantities.** Drag families use the published\n   constant-reference-area force table.\n6. **Flow-structure information.** The image-wake family uses fixed `UxMean`\n   views rather than an integrated coefficient.\n7. **Auditability.** The manifest is distributed with the derived metrics,\n   scripts, figures, LaTeX source, and PDF methods report used to document it.\n\n## Split definitions\n\n### `full`\n\nThe baseline is a seeded random partition over all 500 run identifiers. It\nconstructs `torch.randperm(500)` with seed 42, maps the result to identifiers\n`1..500`, assigns 400 cases to training, 50 to validation, and 50 to testing,\nthen sorts each stored list numerically. The committed identifiers match\nNoether's public `AhmedMLDefaultSplitIDs` assignmen","encoding":"utf8","key":"splits/README.md","name":"README.md","size":8364,"truncated":true},{"bytes":"{\n    \"full_train\": [\n        \"run_1\",\n        \"run_2\",\n        \"run_3\",\n        \"run_5\",\n        \"run_6\",\n        \"run_7\",\n        \"run_8\",\n        \"run_9\",\n        \"run_10\",\n        \"run_13\",\n        \"run_14\",\n        \"run_15\",\n        \"run_16\",\n        \"run_17\",\n        \"run_18\",\n        \"run_21\",\n        \"run_23\",\n        \"run_25\",\n        \"run_27\",\n        \"run_28\",\n        \"run_30\",\n        \"run_31\",\n        \"run_32\",\n        \"run_33\",\n        \"run_34\",\n        \"run_35\",\n        \"run_36\",\n        \"run_37\",\n        \"run_39\",\n        \"run_40\",\n        \"run_42\",\n        \"run_43\",\n        \"run_44\",\n        \"run_45\",\n        \"run_46\",\n        \"run_47\",\n        \"run_48\",\n        \"run_49\",\n        \"run_50\",\n        \"run_51\",\n        \"run_52\",\n        \"run_53\",\n        \"run_54\",\n        \"run_57\",\n        \"run_58\",\n        \"run_60\",\n        \"run_61\",\n        \"run_62\",\n        \"run_63\",\n        \"run_64\",\n        \"run_65\",\n        \"run_66\",\n        \"run_67\",\n        \"run_68\",\n        \"run_69\",\n        \"run_70\",\n        \"run_71\",\n        \"run_72\",\n        \"run_73\",\n        \"run_74\",\n        \"run_75\",\n        \"run_76\",\n        \"run_77\",\n        \"run_78\",\n        \"run_79\",\n        \"run_80\",\n        \"run_81\",\n        \"run_82\",\n        \"run_83\",\n        \"run_84\",\n        \"run_85\",\n        \"run_86\",\n        \"run_87\",\n        \"run_88\",\n        \"run_89\",\n        \"run_90\",\n        \"run_91\",\n        \"run_92\",\n        \"run_93\",\n        \"run_94\",\n        \"run_95\",\n        \"run_96\",\n        \"run_97\",\n        \"run_98\",\n        \"run_99\",\n        \"run_100\",\n        \"run_101\",\n        \"run_102\",\n        \"run_103\",\n        \"run_105\",\n        \"run_106\",\n        \"run_107\",\n        \"run_110\",\n        \"run_111\",\n        \"run_112\",\n        \"run_113\",\n        \"run_114\",\n        \"run_115\",\n        \"run_116\",\n        \"run_117\",\n        \"run_118\",\n        \"run_119\",\n        \"run_120\",\n        \"run_121\",\n        \"run_122\",\n        \"run_123\",\n        \"run_125\",\n        \"run_126\",\n        \"run_128\",\n        \"run_129\",\n        \"run_130\",\n        \"run_131\",\n        \"run_132\",\n        \"run_134\",\n        \"run_135\",\n        \"run_136\",\n        \"run_137\",\n        \"run_138\",\n        \"run_139\",\n        \"run_140\",\n        \"run_141\",\n        \"run_143\",\n        \"run_144\",\n        \"run_145\",\n        \"run_146\",\n        \"run_147\",\n        \"run_148\",\n        \"run_149\",\n        \"run_151\",\n        \"run_152\",\n        \"run_153\",\n        \"run_154\",\n        \"run_155\",\n        \"run_156\",\n        \"run_157\",\n        \"run_159\",\n        \"run_160\",\n        \"run_161\",\n        \"run_162\",\n        \"run_163\",\n        \"run_164\",\n        \"run_166\",\n        \"run_167\",\n        \"run_168\",\n        \"run_169\",\n        \"run_170\",\n        \"run_171\",\n        \"run_172\",\n        \"run_174\",\n        \"run_175\",\n        \"run_176\",\n        \"run_178\",\n        \"run_179\",\n        \"run_181\",\n        \"run_182\",\n        \"run_183\",\n        \"run_184\",\n        \"run_185\",\n        \"run_186\",\n        \"run_189\",\n        \"run_190\",\n        \"run_192\",\n        \"run_193\",\n        \"run_194\",\n        \"run_195\",\n        \"run_198\",\n        \"run_200\",\n        \"run_201\",\n        \"run_202\",\n        \"run_204\",\n        \"run_206\",\n        \"run_209\",\n        \"run_211\",\n        \"run_212\",\n        \"run_213\",\n        \"run_214\",\n        \"run_216\",\n        \"run_217\",\n        \"run_218\",\n        \"run_219\",\n        \"run_220\",\n        \"run_221\",\n        \"run_223\",\n        \"run_224\",\n        \"run_225\",\n        \"run_227\",\n        \"run_229\",\n        \"run_231\",\n        \"run_232\",\n        \"run_233\",\n        \"run_235\",\n        \"run_236\",\n        \"run_237\",\n        \"run_238\",\n        \"run_239\",\n        \"run_240\",\n        \"run_242\",\n        \"run_243\",\n        \"run_244\",\n        \"run_245\",\n        \"run_246\",\n        \"run_248\",\n        \"run_249\",\n        \"run_250\",\n        \"run_251\",\n        \"run_254\",\n        \"run_255\",\n        \"run_256\",\n        \"run_257\",\n        \"run_259\",\n        \"run_261\",\n        \"run_262\",\n        \"run_264\",\n        \"run_265\",\n        \"run_266\",\n        \"run_267\",\n        \"run_268\",\n        \"run_269\",\n        \"run_270\",\n      ","encoding":"utf8","key":"splits/manifest.json","name":"manifest.json","size":57273,"truncated":true}],"sourceObjectKey":"force_mom_all.csv","starterCell":"import pandas as pd\ndf = pd.read_csv(\"hf://datasets/neashton/ahmedml/force_mom_all.csv\")\ndf.head()","starterSamplePath":"force_mom_all.csv","totalObjects":61528},"summary":"AhmedML contains 500 geometric variations of the Ahmed car body, a simplified car-like shape that reproduces flow features of road vehicles such as geometry- and pressure-induced…","title":"AhmedML: High-Fidelity CFD Dataset for Low-Speed Bluff-Body Aerodynamics","updatedAt":"2026-11-14T17:37:42.845Z","visibility":"public","websiteUrl":"https://caemldatasets.org/ahmedml/"}}