onescience_run_manifest.yaml

onescience_manifest_version: "0.1"
resource_type: dataset
resource:
  id: "OneScience/evo2_dataset"
  name: "Evo2 mini genome dataset"
  domain: bio
  domain_tags:
    - bio
    - genomics
    - dna_sequence
  task: genome_sequence_modeling_data
  task_tags:
    - train_data
    - validation_data
    - test_data
    - preflight
  modalities:
    - dna_sequence
  input_formats:
    - fasta
    - fasta_gzip
    - indexed_dataset_bin_idx
  output_formats: []
  summary: "适配 OneScience/evo2/ 的 chr20/chr21/chr22 mini 基因组数据集，包含 FASTA 原始文件和 Byte-Level 预处理 bin/idx split。"

platform_resource:
  primary:
    platform: modelscope
    repo_id: "OneScience/evo2_dataset"
    repo_type: dataset
    url: "https://modelscope.cn/datasets/OneScience/evo2_dataset"
    revision: main
    readme_path: README.md
    manifest_path: manifest.yaml
  mirrors: []
  access:
    visibility: public
    license: "Dataset license follows source genome data terms"

website_integration:
  enabled: true
  click_target:
    platform: modelscope
    resource_url: "https://modelscope.cn/datasets/OneScience/evo2_dataset"
  llm_handoff:
    readme_required: true
    manifest_required: true
    download_readme_first: true
    resolve_related_models: true
    default_run_goal: minimal_validation
    cwd_rule: "如果使用 modelscope download --cache_dir 下载，运行 cwd 必须切换到实际下载后的数据集包根目录，即包含 README.md 和 manifest.yaml 的目录。"

runtime:
  enabled: true
  onescience_domain: bio
  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: "pip install -c constraints.txt .[bio]"
  source_paths: []
  compatibility:
    examples_path: "examples/biosciences/evo2"
    status: examples_compatible

runtime_package:
  kind: standard_runtime_package
  package_root: .
  standard_layout:
    workdir: .
    config_dir: null
    checkpoint_dir: null
    data_dir: data_mini
    output_dir: outputs
  apply_policy:
    mode: direct_use
    target: session_workdir
    overwrite: true
    protect_installed_onescience: true
  entry_files:
    - scripts/validate_evo2_dataset.py
    - data_mini/genome_data/
    - metadata/data_mini.files.tsv
    - metadata/data_mini.sha256
  entrypoints:
    preflight: scripts/validate_evo2_dataset.py
    inference: null
    train: null
    finetune: null
    evaluate: null
    visualize: null
    deploy: null

files:
  model_files: []
  dataset_files:
    - id: data_mini
      path: data_mini/
      role: train_eval_test_data
      format: fasta_and_indexed_dataset
      description_zh: "Evo2 mini 数据根目录，包含 genome_data 下的 chr20/chr21/chr22 FASTA、压缩 FASTA、合并 FASTA 和预处理 bin/idx split。"
      required: true
      required_for:
        - preflight
        - train
        - predict
      source:
        platform: modelscope
        repo_id: "OneScience/evo2_dataset"
        repo_type: dataset
        path: data_mini/
        revision: main
        download_method: command_ref
        command_ref: commands.download.download_dataset
      local_path: data_mini/
      checksum_manifest:
        sha256: metadata/data_mini.sha256
        sizes: metadata/data_mini.files.tsv
    - id: genome_fasta
      path: data_mini/genome_data/
      role: raw_fasta
      format: fasta
      description_zh: "chr20.fa、chr21.fa、chr22.fa 和 chr20_21_22.fa 原始序列文件。"
      required: true
      required_for:
        - preflight
        - preprocess
        - predict
      source:
        platform: modelscope
        repo_id: "OneScience/evo2_dataset"
        repo_type: dataset
        path: data_mini/genome_data/
        revision: main
        download_method: command_ref
        command_ref: commands.download.download_dataset
      local_path: data_mini/genome_data/
    - id: indexed_splits
      path: data_mini/genome_data/preprocessed_data/
      role: indexed_dataset
      format: bin_idx
      description_zh: "训练、验证、测试 split 的 chr20_21_22_uint8_distinct_byte-level_*.bin 和 *.idx。"
      required: true
      required_for:
        - preflight
        - train
      source:
        platform: modelscope
        repo_id: "OneScience/evo2_dataset"
        repo_type: dataset
        path: data_mini/genome_data/preprocessed_data/
        revision: main
        download_method: command_ref
        command_ref: commands.download.download_dataset
      local_path: data_mini/genome_data/preprocessed_data/
  config_files: []
  sample_files: []

dataset:
  format:
    - fasta
    - fasta_gzip
    - bin_idx
  sample_unit: "DNA 序列片段或 Byte-Level indexed dataset 样本。"
  schema:
    path: metadata/data_mini.files.tsv
    fields:
      - relative_path
      - size_bytes
      - sha256
  splits:
    train: data_mini/genome_data/preprocessed_data/chr20_21_22_uint8_distinct_byte-level_train
    validation: data_mini/genome_data/preprocessed_data/chr20_21_22_uint8_distinct_byte-level_val
    test: data_mini/genome_data/preprocessed_data/chr20_21_22_uint8_distinct_byte-level_test

relations:
  required_datasets: []
  optional_datasets: []
  compatible_models:
    - id: "OneScience/evo2/"
      role: train_data
      required_for:
        - preflight
        - train
        - predict
      resource_ref:
        platform: modelscope
        repo_id: "OneScience/evo2/"
        repo_type: model
        url: "https://modelscope.cn/models/OneScience/evo2/"
        revision: main
        readme_path: README.md
        manifest_path: manifest.yaml
      expected_local_path: data/evo2_dataset/data_mini/genome_data/
      adapter: "模型包 config/genome_data_config.yaml 和 config/opengenome2.yml"

run_matrix:
  scenarios:
    - name: minimal_validation
      default: true
      capability: preflight
      description_zh: "校验数据集目录结构、FASTA 可读性、bin/idx split、文件大小和 SHA256。"
      required_dataset_files:
        - data_mini
        - genome_fasta
        - indexed_splits
      required_model_files: []
      required_datasets: []
      preconditions:
        - "已下载数据集仓库 OneScience/evo2_dataset。"
        - "cwd 位于数据集包根目录。"
      command_refs:
        - commands.preflight.validate_dataset
      outputs:
        - "终端输出 [OK] Evo2 mini dataset..."
    - name: model_train_input
      capability: train
      description_zh: "作为 OneScience/evo2/ 训练或微调输入数据。"
      required_dataset_files:
        - indexed_splits
      required_model_files: []
      required_datasets: []
      preconditions:
        - "将 data_mini 复制或链接到模型包 data/evo2_dataset/data_mini。"
      command_refs:
        - commands.prepare.place_for_model
      outputs:
        - "<MODEL_REPO_ROOT>/data/evo2_dataset/data_mini"

capabilities:
  inference: false
  train: true
  finetune: true
  evaluate: false
  visualize: false
  deploy: false
  preflight: true
  data_reading: true

commands:
  download:
    - name: download_dataset
      cwd: .
      command: "modelscope download --dataset OneScience/evo2_dataset"
      description_zh: "下载数据集标准包。若使用 --cache_dir，下载完成后必须 cd 到包含 README.md 和 manifest.yaml 的数据集包根目录。"
    - name: download_model
      cwd: .
      command: "modelscope download --model OneScience/evo2/"
      description_zh: "下载兼容模型 OneScience/evo2/。"
  prepare:
    - name: place_for_model
      cwd: .
      command: "mkdir -p <MODEL_REPO_ROOT>/data/evo2_dataset && cp -a data_mini <MODEL_REPO_ROOT>/data/evo2_dataset/data_mini"
      description_zh: "把数据集放到模型包默认数据路径。"
  preflight:
    - name: validate_dataset
      cwd: .
      command: "python scripts/validate_evo2_dataset.py --package-root . --dataset-root data_mini"
      description_zh: "检查数据结构、FASTA 可读性、预处理 split、大小和 SHA256。"
  inference: []
  train: []
  finetune: []
  evaluate: []
  visualize: []
  deploy: []

expected_outputs:
  - id: dataset_validation_ok
    path: stdout
    description_zh: "数据集验证成功时输出 [OK] Evo2 mini dataset..."
    produced_by: commands.preflight.validate_dataset
  - id: placed_dataset
    path: "<MODEL_REPO_ROOT>/data/evo2_dataset/data_mini"
    description_zh: "供 OneScience/evo2/ 默认预检和训练命令读取的数据目录。"
    produced_by: commands.prepare.place_for_model

diagnostics:
  - id: missing_fasta
    symptom: "missing FASTA file"
    likely_cause_zh: "data_mini 未完整下载或目录位置不对。"
    fix_zh: "重新运行 modelscope download --dataset OneScience/evo2_dataset，或用 --dataset-root 指向实际 data_mini。"
  - id: missing_bin_idx
    symptom: "missing preprocessed bin or idx"
    likely_cause_zh: "预处理 split 文件缺失。"
    fix_zh: "重新下载数据集，确认 data_mini/genome_data/preprocessed_data 下有 train/val/test 的 .bin 和 .idx。"
  - id: sha256_mismatch
    symptom: "sha256 mismatch"
    likely_cause_zh: "上传、下载或复制过程中数据损坏。"
    fix_zh: "删除损坏文件后重新下载 OneScience/evo2_dataset。"
  - id: model_path_mismatch
    symptom: "model preflight cannot find data/evo2_dataset/data_mini"
    likely_cause_zh: "数据集没有放到模型包预期路径。"
    fix_zh: "在模型包根目录执行 mkdir -p data/evo2_dataset && cp -a <DATASET_REPO_ROOT>/data_mini data/evo2_dataset/data_mini。"

domain_extension:
  bio:
    organism_scope_zh: "人类 chr20、chr21、chr22 mini 基因组数据。"
    sequence_alphabet: "DNA / Byte-Level tokenizer"
    raw_files:
      - data_mini/genome_data/chr20.fa
      - data_mini/genome_data/chr21.fa
      - data_mini/genome_data/chr22.fa
      - data_mini/genome_data/chr20_21_22.fa
    indexed_dataset_dtype: uint8
    preprocessing_config_in_model: "OneScience/evo2/ config/genome_preprocess_config.yaml"