OneScience-Group/cfd_benchmark

1onescience_manifest_version: "0.1"2resource_type: dataset3resource:4  id: OneScience/cfd_benchmark5  name: cfd_benchmark6  domain: cfd7  domain_tags:8    - cfd9    - neural_pde_solver10    - benchmark11  task: cfd_operator_learning_dataset12  task_tags:13    - flow_field_prediction14    - pde_operator_learning15    - data_reading_validation16  modalities:17    - gridded_field18    - unstructured_mesh19  input_formats:20    - npy21    - mat22  output_formats: []23  summary: CFD_Benchmark 数据集的 ModelScope 标准数据包，本次包含 airfoil、darcy、elasticity、ns、pipe、plas 六个子集。24 25platform_resource:26  primary:27    platform: modelscope28    repo_id: OneScience/cfd_benchmark29    repo_type: dataset30    url: https://modelscope.cn/datasets/OneScience/cfd_benchmark31    revision: main32    readme_path: README.md33    manifest_path: manifest.yaml34  mirrors: []35  access:36    visibility: public37    license: null38 39website_integration:40  enabled: true41  click_target:42    platform: modelscope43    resource_url: https://modelscope.cn/datasets/OneScience/cfd_benchmark44  llm_handoff:45    readme_required: true46    manifest_required: true47    download_readme_first: true48    resolve_related_models: true49    default_run_goal: minimal_validation50 51runtime:52  enabled: true53  onescience_domain: cfd54  min_onescience_version: null55  supported_execution:56    - local57    - local_slurm58    - remote_slurm59 60onescience:61  repo: https://gitee.com/onescience-ai/onescience62  official_links:63    gitee:64      doc: https://gitee.com/onescience-ai/onescience-doc65      onescience: https://gitee.com/onescience-ai/onescience66      skills: https://gitee.com/onescience-ai/oneskills67    github:68      doc: https://github.com/onescience-ai/OneScience-doc69      onescience: https://github.com/onescience-ai/OneScience70      skills: https://github.com/onescience-ai/oneskills71  install:72    required_by_default: false73    command: source ~/env.sh && bash install.sh cfd74  source_paths: []75  compatibility:76    examples_path: examples/cfd/CFD_Benchmark77    status: examples_compatible78 79runtime_package:80  kind: standard_runtime_package81  package_root: .82  standard_layout:83    workdir: .84    config_dir: metadata85    checkpoint_dir: null86    data_dir: data87    output_dir: reports88  apply_policy:89    mode: direct_use90    target: session_workdir91    overwrite: true92    protect_installed_onescience: true93  entry_files:94    - metadata/file_manifest.csv95    - metadata/schema.yaml96    - scripts/validate_dataset.py97  entrypoints:98    preflight: scripts/validate_dataset.py99    inference: null100    train: null101    finetune: null102    evaluate: null103    visualize: null104    deploy: null105 106files:107  model_files: []108  dataset_files:109    - id: cfd_benchmark_data110      path: data/111      role: train_data112      description_zh: 本次整理的 CFD_Benchmark 六个子集数据目录，包含 airfoil、darcy、elasticity、ns、pipe、plas。113      required: true114      required_for:115        - train116        - evaluate117        - preflight118      format: npy_mat119      source:120        platform: modelscope121        repo_id: OneScience/cfd_benchmark122        repo_type: dataset123        path: data/124        revision: main125        download_method: command_ref126        command_ref: commands.download.download_dataset127      local_path: session_workdir/cfd_cfd_benchmark_data/data/128      size: "约 8.2G"129      checksum: metadata/file_manifest.csv130      large_file: true131    - id: file_manifest132      path: metadata/file_manifest.csv133      role: checksum_manifest134      description_zh: 整理后每个数据文件的相对路径、字节数和 SHA256。135      required: true136      required_for:137        - preflight138      format: csv139      source:140        platform: modelscope141        repo_id: OneScience/cfd_benchmark142        repo_type: dataset143        path: metadata/file_manifest.csv144        revision: main145        download_method: command_ref146        command_ref: commands.download.download_dataset147      local_path: session_workdir/cfd_cfd_benchmark_data/metadata/file_manifest.csv148      size: null149      checksum: null150      large_file: false151  config_files:152    - id: schema153      path: metadata/schema.yaml154      role: schema155      description_zh: 记录 npy shape/dtype 和 MATLAB 文件变量结构。156      required: true157      required_for:158        - preflight159        - train160        - evaluate161      format: yaml162      source:163        platform: modelscope164        repo_id: OneScience/cfd_benchmark165        repo_type: dataset166        path: metadata/schema.yaml167        revision: main168        download_method: command_ref169        command_ref: commands.download.download_dataset170      local_path: session_workdir/cfd_cfd_benchmark_data/metadata/schema.yaml171      checksum: null172      large_file: false173  sample_files: []174 175dataset:176  format:177    - npy178    - matlab_v5_mat179  sample_unit: 不同子集的一个样本定义不同，airfoil/pipe 为一个几何流场样本，darcy/ns/plas 为一个 PDE 系数或时间序列样本，elasticity 为一个随机单元结构样本。180  schema:181    path: metadata/schema.yaml182  checksum_manifest:183    path: metadata/file_manifest.csv184  splits:185    train: 由 CFD_Benchmark 运行脚本通过 ntrain 或 train_ratio 切分。186    validation: null187    test: 由 CFD_Benchmark 运行脚本通过 ntest 或 train_ratio 切分。188 189relations:190  required_datasets: []191  optional_datasets: []192  compatible_models:193    - id: OneScience/CFD_Benchmark194      role: train_data195      required_for:196        - train197        - evaluate198        - preflight199      resource_ref:200        platform: modelscope201        repo_id: OneScience/CFD_Benchmark202        repo_type: model203        url: https://modelscope.cn/models/OneScience/CFD_Benchmark204        revision: main205        readme_path: README.md206        manifest_path: manifest.yaml207      expected_local_path: ../cfd_cfd_benchmark_data/data208      adapter: scripts/standard_train_airfoil_transolver.sh209 210run_matrix:211  scenarios:212    - name: minimal_validation213      capability: preflight214      default: true215      required_datasets:216        - id: OneScience/cfd_benchmark217          role: train_data218          local_path: data/219      required_model_files: []220      command_refs:221        - commands.preflight.validate_dataset222      expected_outputs:223        - reports/dataset_validation.json224      prerequisites:225        - 已下载数据集 OneScience/cfd_benchmark。226    - name: cfd_benchmark_model_train227      capability: train228      required_datasets:229        - id: OneScience/cfd_benchmark230          role: train_data231          local_path: data/232      required_model_files:233        - OneScience/CFD_Benchmark runtime/234      command_refs:235        - commands.preflight.validate_dataset236      expected_outputs:237        - reports/dataset_validation.json238      prerequisites:239        - 下载并进入模型仓库 OneScience/CFD_Benchmark。240 241capabilities:242  inference: false243  train: false244  finetune: false245  evaluate: false246  visualize: false247  deploy: false248  data_reading_validation: true249 250commands:251  download:252    - name: download_dataset253      description: 下载 CFD_Benchmark 数据集标准包。254      cwd: .255      command: modelscope download --dataset OneScience/cfd_benchmark --local_dir ./cfd_cfd_benchmark_data256      required_files: []257      required_env:258        - MODELSCOPE_TOKEN259      expected_outputs:260        - path: cfd_cfd_benchmark_data/manifest.yaml261          type: file262        - path: cfd_cfd_benchmark_data/data/263          type: directory264      success_criteria:265        - exit_code == 0266  preflight:267    - name: validate_dataset268      description: 校验 14 个数据文件的存在性、大小、SHA256、npy shape/dtype 和 mat 变量结构。269      cwd: .270      command: source ~/env.sh && python scripts/validate_dataset.py --data-root data --file-manifest metadata/file_manifest.csv271      required_files:272        - scripts/validate_dataset.py273        - metadata/file_manifest.csv274        - data/275      required_env:276        - python277        - numpy278        - scipy279      expected_outputs:280        - path: reports/dataset_validation.json281          type: file282      success_criteria:283        - exit_code == 0284  prepare: []285  inference: []286  train: []287  finetune: []288  evaluate: []289  visualize: []290  deploy: []291 292expected_outputs:293  - path: reports/dataset_validation.json294    type: json295    description_zh: 数据文件完整性和可读性验证报告。296  - path: metadata/file_manifest.csv297    type: csv298    description_zh: 文件名、大小和 SHA256 校验清单。299  - path: metadata/schema.yaml300    type: yaml301    description_zh: 数据结构说明。302 303diagnostics:304  - id: missing_file305    symptom: validate_dataset.py 报告 status=missing。306    cause: 数据下载不完整或放置目录错误。307    fix: 重新执行 modelscope download --dataset OneScience/cfd_benchmark --local_dir ./cfd_cfd_benchmark_data。308  - id: sha256_mismatch309    symptom: validate_dataset.py 报告 status=sha256_mismatch。310    cause: 文件损坏、复制中断或上传下载过程中内容变化。311    fix: 删除对应文件后重新下载数据集，或用 metadata/source_file_manifest.csv 与源目录重新比对。312  - id: scipy_missing313    symptom: MATLAB 文件只完成 header_only 检查。314    cause: 当前 Python 环境缺少 scipy。315    fix: source ~/env.sh 后重跑验证；当前标准化过程中已用 scipy 生成 metadata/schema.yaml。316 317domain_extension:318  cfd:319    benchmark_name: CFD_Benchmark320    included_subsets:321      airfoil:322        files:323          - NACA_Cylinder_X.npy324          - NACA_Cylinder_Y.npy325          - NACA_Cylinder_Q.npy326      darcy:327        files:328          - piececonst_r421_N1024_smooth1.mat329          - piececonst_r421_N1024_smooth2.mat330      elasticity:331        files:332          - Meshes/Random_UnitCell_XY_10.npy333          - Meshes/Random_UnitCell_rr_10.npy334          - Meshes/Random_UnitCell_sigma_10.npy335          - Meshes/Random_UnitCell_theta_10.npy336      ns:337        files:338          - NavierStokes_V1e-5_N1200_T20.mat339      pipe:340        files:341          - Pipe_X.npy342          - Pipe_Y.npy343          - Pipe_Q.npy344      plas:345        files:346          - plas_N987_T20.mat347    skipped_source_subsets:348      - car349      - diffusion_sorption350