onescience_run_manifest.yaml

onescience_manifest_version: "0.1"
resource_type: dataset
resource:
  id: OneScience/water
  name: water
  domain: matchem
  domain_tags:
    - matchem
    - water
    - atomistic_dataset
  task: mlip_training_data
  task_tags:
    - mlip
    - energy_prediction
    - force_prediction
    - extxyz_training
  modalities:
    - atomistic_structure
  input_formats:
    - extxyz
  output_formats: []
  summary: 面向 MACE 水体系势能面训练和评测示例的 extxyz 数据集。

platform_resource:
  primary:
    platform: modelscope
    repo_id: OneScience/water
    repo_type: dataset
    url: https://modelscope.cn/datasets/OneScience/water
    revision: main
    readme_path: README.md
    manifest_path: manifest.yaml
  mirrors: []
  access:
    visibility: public
    license: null

website_integration:
  enabled: true
  click_target:
    platform: modelscope
    resource_url: https://modelscope.cn/datasets/OneScience/water
  llm_handoff:
    readme_required: true
    manifest_required: true
    download_readme_first: true
    resolve_related_models: true
    default_run_goal: minimal_validation

runtime:
  enabled: true
  onescience_domain: matchem
  min_onescience_version: null
  supported_execution:
    - local
    - local_slurm
    - remote_slurm
    - scnet_mcp

onescience:
  repo: https://gitee.com/onescience-ai/onescience
  official_links:
    gitee:
      doc: https://gitee.com/onescience-ai/onescience-doc
      onescience: https://gitee.com/onescience-ai/onescience
      skills: https://gitee.com/onescience-ai/oneskills
    github:
      doc: https://github.com/onescience-ai/OneScience-doc
      onescience: https://github.com/onescience-ai/OneScience
      skills: https://github.com/onescience-ai/oneskills
  install:
    required_by_default: false
    command: bash install.sh matchem
  source_paths:
    - matchem/mace/water
  compatibility:
    examples_path: examples/matchem/mace
    status: examples_compatible

runtime_package:
  kind: standard_runtime_package
  package_root: .
  standard_layout:
    workdir: .
    config_dir: metadata
    checkpoint_dir: null
    data_dir: data
    output_dir: outputs
  apply_policy:
    mode: direct_use
    target: session_workdir
    overwrite: true
    protect_installed_onescience: true
  entry_files:
    - scripts/validate_water.py
    - metadata/schema.yaml
    - metadata/sha256_manifest.txt
  entrypoints:
    preflight: scripts/validate_water.py
    inference: null
    train: null
    finetune: null
    evaluate: null
    visualize: null
    deploy: null

derivation:
  source_file: data/water/dataset_1593.xyz
  generated_files:
    - path: data/water/water_train.xyz
      method: frames[0:1393]
      reason: MACE water 配置读取 water_train.xyz；原始目录只有 dataset_1593.xyz，因此按 gene_xyz.py 逻辑派生。
    - path: data/water/water_test.xyz
      method: frames[1393:1593]
      reason: MACE water 配置读取 water_test.xyz；原始目录只有 dataset_1593.xyz，因此按 gene_xyz.py 逻辑派生。
  verification: water_train.xyz 与 water_test.xyz 字节级拼接后等于 dataset_1593.xyz。

files:
  model_files: []
  dataset_files:
    - id: water_full_extxyz
      path: data/water/dataset_1593.xyz
      role: metadata
      description_zh: 原始 water extxyz 全量文件，包含 1593 帧，每帧 192 个原子。
      required: true
      required_for: [preflight]
      format: extxyz
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: data/water/dataset_1593.xyz
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/data/water/dataset_1593.xyz
      size: 34M
      checksum: metadata/sha256_manifest.txt
      large_file: false
    - id: water_train_extxyz
      path: data/water/water_train.xyz
      role: train_split
      description_zh: 从 dataset_1593.xyz 前 1393 帧派生的 MACE 训练文件。
      required: true
      required_for: [preflight, train]
      format: extxyz
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: data/water/water_train.xyz
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/data/water/water_train.xyz
      size: 30M
      checksum: metadata/sha256_manifest.txt
      large_file: false
    - id: water_test_extxyz
      path: data/water/water_test.xyz
      role: test_split
      description_zh: 从 dataset_1593.xyz 后 200 帧派生的 MACE 测试文件。
      required: true
      required_for: [preflight, evaluate]
      format: extxyz
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: data/water/water_test.xyz
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/data/water/water_test.xyz
      size: 4.3M
      checksum: metadata/sha256_manifest.txt
      large_file: false
    - id: water_generation_script
      path: data/water/gene_xyz.py
      role: metadata
      description_zh: 原始目录提供的 split 生成脚本，说明 train/test 的切分方式。
      required: false
      required_for: [preflight]
      format: py
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: data/water/gene_xyz.py
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/data/water/gene_xyz.py
      size: 236B
      checksum: metadata/sha256_manifest.txt
      large_file: false
  config_files:
    - id: schema
      path: metadata/schema.yaml
      role: schema
      description_zh: water extxyz 数据结构、字段名、shape、单位、split 和 MACE 配置键说明。
      required: true
      required_for: [preflight, train, evaluate]
      format: yaml
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: metadata/schema.yaml
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/metadata/schema.yaml
      size: null
      checksum: null
      large_file: false
    - id: sha256_manifest
      path: metadata/sha256_manifest.txt
      role: checksum
      description_zh: water 数据文件 SHA256 清单，用于校验整理后数据与原始数据及派生 split 一致。
      required: true
      required_for: [preflight]
      format: txt
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: metadata/sha256_manifest.txt
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/metadata/sha256_manifest.txt
      size: null
      checksum: null
      large_file: false
  sample_files:
    - id: validate_script
      path: scripts/validate_water.py
      role: script
      description_zh: water 数据读取验证脚本，检查 extxyz 字段、shape、帧数、元素集合、SHA256 和 split 还原。
      required: true
      required_for: [preflight]
      format: py
      source:
        platform: modelscope
        repo_id: OneScience/water
        repo_type: dataset
        path: scripts/validate_water.py
        revision: main
        url: null
        download_method: command_ref
        command_ref: commands.download.download_dataset_package
      mirrors: []
      local_path: session_workdir/scripts/validate_water.py
      size: null
      checksum: null
      large_file: false

dataset:
  format:
    - extxyz
  sample_unit: 单个水体系构型，包含元素、坐标、TotEnergy、force 和 Z。
  schema:
    path: metadata/schema.yaml
  splits:
    full:
      path: data/water/dataset_1593.xyz
      frames: 1593
      atoms_per_frame: 192
    train:
      path: data/water/water_train.xyz
      frames: 1393
      atoms_per_frame: 192
    validation:
      path: null
      note_zh: "MACE water 示例通过 valid_fraction: 0.05 从训练文件划分验证集。"
    test:
      path: data/water/water_test.xyz
      frames: 200
      atoms_per_frame: 192

relations:
  required_datasets: []
  optional_datasets: []
  compatible_models:
    - id: OneScience/MACE
      role: train_data
      required_for: [preflight, train, evaluate]
      resource_ref:
        platform: modelscope
        repo_id: OneScience/MACE
        repo_type: model
        url: https://modelscope.cn/models/OneScience/MACE
        revision: main
        readme_path: README.md
        manifest_path: manifest.yaml
      expected_local_path: session_workdir/data/water
      adapter: null

run_matrix:
  scenarios:
    - name: validate_dataset
      default: true
      capability: preflight
      required_datasets:
        - id: OneScience/water
          role: train_data
          local_path: data/water
      required_model_files: []
      required_data_files:
        - data/water/dataset_1593.xyz
        - data/water/water_train.xyz
        - data/water/water_test.xyz
        - metadata/sha256_manifest.txt
      preconditions:
        - 已下载数据集仓库 OneScience/water。
      command_refs:
        - commands.preflight.validate_dataset
      outputs:
        - path: stdout
          description_zh: 数据字段、帧数、split 还原和 SHA256 检查结果。

capabilities:
  inference: false
  train: false
  finetune: false
  evaluate: false
  visualize: false
  deploy: false

commands:
  download:
    - name: download_dataset_package
      target: dataset
      repo_id: OneScience/water
      source: .
      local_path: session_workdir/
      required: true
      required_for: [preflight, train, evaluate]
      cwd: .
      command: modelscope download --dataset OneScience/water --local_dir session_workdir
      expected_outputs:
        - path: session_workdir/data/water/water_train.xyz
          type: file
      success_criteria:
        - exit_code == 0
        - output_path_exists
  preflight:
    - name: validate_dataset
      description: 验证 water extxyz 文件结构、字段名、帧数、split 还原和 SHA256。
      cwd: .
      command: python scripts/validate_water.py --dataset-root data/water --checksum-manifest metadata/sha256_manifest.txt
      required_files:
        - data/water/dataset_1593.xyz
        - data/water/water_train.xyz
        - data/water/water_test.xyz
        - metadata/sha256_manifest.txt
        - scripts/validate_water.py
      required_env:
        - python
      expected_outputs:
        - path: stdout
          type: text
      success_criteria:
        - exit_code == 0
  prepare: []
  inference: []
  train: []
  finetune: []
  evaluate: []
  visualize: []
  deploy: []

expected_outputs:
  - path: stdout
    type: text
    description_zh: 验证脚本输出每个 split、字段、split 还原和 SHA256 的 OK 结果。

diagnostics:
  - error_pattern: "缺少文件"
    likely_cause: 数据集下载不完整或运行 cwd 不在数据集仓库根目录。
    action: 重新执行 modelscope download --dataset OneScience/water，并在下载后的数据集根目录运行验证命令。
  - error_pattern: "SHA256 不一致"
    likely_cause: 文件被截断、损坏或被替换。
    action: 删除损坏文件后重新下载数据集。
  - error_pattern: "TotEnergy"
    likely_cause: extxyz comment 中缺少 MACE water 配置要求的能量字段。
    action: 检查数据是否为本仓库 water 数据，或同步修改模型配置 energy_key。
  - error_pattern: "force"
    likely_cause: extxyz Properties 中缺少 MACE water 配置要求的力字段。
    action: 检查数据是否为本仓库 water 数据，或同步修改模型配置 forces_key。
  - error_pattern: "不能字节级还原"
    likely_cause: 派生的 water_train.xyz 或 water_test.xyz 与 dataset_1593.xyz 的切分方式不一致。
    action: 按 gene_xyz.py 重新从 dataset_1593.xyz 生成前 1393 帧训练和后 200 帧测试。

domain_extension:
  matchem:
    chemical_system:
      elements:
        - H
        - O
    properties:
      - energy
      - forces
    units:
      energy: eV
      forces: eV/Angstrom
    structure_format:
      - extxyz

large_resource:
  enabled: false
  total_size: 68M
  file_count: 4
  download_policy: full
  default_download: full
  minimal_files:
    - data/water/water_train.xyz
    - data/water/water_test.xyz
    - data/water/dataset_1593.xyz

reproducibility:
  level: minimal
  goal: 完成一次 water extxyz 数据读取、split 还原和 SHA256 校验。
  command_ref: commands.preflight.validate_dataset
  expected_runtime: 1-2 minutes
  expected_outputs:
    - stdout