onescience_manifest_version: "0.1" resource_type: dataset resource: id: OneScience/lagrangian name: lagrangian domain: cfd domain_tags: [cfd, particle_simulation, lagrangian_dynamics] task: lagrangian_particle_simulation_dataset task_tags: [train_data, eval_data, inference_input, dataset_validation] modalities: [particle_trajectory] input_formats: [tfrecord] output_formats: [tfrecord] summary: DeepMind Lagrangian Water 子数据集标准数据包，包含 train、valid、test 三个 TFRecord split 和 metadata.json。 platform_resource: primary: platform: modelscope repo_id: OneScience/lagrangian repo_type: dataset url: https://modelscope.cn/datasets/OneScience/lagrangian revision: main readme_path: README.md manifest_path: onescience_run_manifest.yaml mirrors: [] access: visibility: public license: unknown website_integration: enabled: true click_target: platform: modelscope resource_url: https://modelscope.cn/datasets/OneScience/lagrangian llm_handoff: readme_required: true manifest_required: true download_readme_first: true resolve_related_models: true default_run_goal: dataset_validation cwd_note: 如果使用 modelscope download --cache_dir 下载数据集，请先 cd 到实际下载后的数据集仓库根目录；模型侧将 ONESCIENCE_LAGRANGIAN_DATA_DIR 指向本仓库 data/Water。 runtime: enabled: true onescience_domain: cfd min_onescience_version: null supported_execution: [local_dataset_validation, model_training_input, model_inference_input] environment: exported_env: ONESCIENCE_LAGRANGIAN_DATA_DIR: /data/Water dependencies: python: ">=3.10" python_packages: [tensorflow, numpy, pyyaml] 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 cfd source_paths: - onescience/src/onescience/datapipes/cfd/deepmind_lagrangian.py compatibility: examples_path: onescience/examples/cfd/Lagrangian_MGN status: examples_compatible datapipe: onescience.datapipes.cfd.DeepMindLagrangianDatapipe runtime_package: kind: standard_runtime_package package_root: . standard_layout: workdir: . data_dir: data/Water metadata_dir: metadata output_dir: validation_outputs apply_policy: mode: direct_use target: session_workdir overwrite: false protect_installed_onescience: true entry_files: - data/Water/metadata.json - data/Water/train.tfrecord - data/Water/valid.tfrecord - data/Water/test.tfrecord - metadata/lagrangian_water_schema.json - scripts/validate_lagrangian_dataset.py - files_sha256.jsonl entrypoints: preflight: scripts/validate_lagrangian_dataset.py validate: scripts/validate_lagrangian_dataset.py inference: null train: null finetune: null evaluate: null visualize: null deploy: null files: model_files: [] dataset_files: - id: water_tfrecord_splits path: data/Water/*.tfrecord role: train_eval_inference_data description_zh: Water 子数据集的 train、valid、test TFRecord 序列文件。 required: true required_for: [dataset_validation, train, inference, evaluate] source: {platform: modelscope, repo_id: OneScience/lagrangian, repo_type: dataset, download_method: command_ref, command_ref: commands.download.download_dataset} local_path: data/Water - id: water_metadata path: data/Water/metadata.json role: schema_and_normalization_stats description_zh: Water 数据的维度、时间步长、边界、半径、速度和加速度归一化统计。 required: true required_for: [dataset_validation, train, inference, evaluate] source: {platform: modelscope, repo_id: OneScience/lagrangian, repo_type: dataset, download_method: command_ref, command_ref: commands.download.download_dataset} local_path: data/Water/metadata.json - id: integrity_inventory path: files_sha256.jsonl role: file_size_sha256_inventory description_zh: 整理后数据文件大小和 SHA256 清单。 required: true required_for: [dataset_validation] source: {platform: modelscope, repo_id: OneScience/lagrangian, repo_type: dataset, download_method: command_ref, command_ref: commands.download.download_dataset} local_path: files_sha256.jsonl - id: integrity_summary path: data_integrity_summary.json role: source_to_standardized_integrity_summary description_zh: 原始数据与整理后数据的文件名、大小、SHA256 对照摘要。 required: true required_for: [dataset_validation] source: {platform: modelscope, repo_id: OneScience/lagrangian, repo_type: dataset, download_method: command_ref, command_ref: commands.download.download_dataset} local_path: data_integrity_summary.json config_files: - id: schema path: metadata/lagrangian_water_schema.json role: dataset_schema description_zh: TFRecord SequenceExample 字段、dtype、shape 和 split 数量说明。 required: true required_for: [dataset_validation, train, inference] source: {platform: modelscope, repo_id: OneScience/lagrangian, repo_type: dataset, download_method: command_ref, command_ref: commands.download.download_dataset} local_path: metadata/lagrangian_water_schema.json sample_files: - id: validation_script path: scripts/validate_lagrangian_dataset.py role: dataset_validation description_zh: 检查数据文件结构、metadata、TFRecord 首条样本 shape/dtype 和可选 SHA256。 required: true required_for: [dataset_validation] source: {platform: modelscope, repo_id: OneScience/lagrangian, repo_type: dataset, download_method: command_ref, command_ref: commands.download.download_dataset} local_path: scripts/validate_lagrangian_dataset.py dataset: format: [tfrecord_sequence_example] sample_unit: 一个 SequenceExample 对应一条粒子仿真序列。 schema: path: metadata/lagrangian_water_schema.json splits: train: {path: data/Water/train.tfrecord, expected_sequences: 1000} validation: {path: data/Water/valid.tfrecord, expected_sequences: 30} test: {path: data/Water/test.tfrecord, expected_sequences: 30} metadata: path: data/Water/metadata.json dim: 2 sequence_length: 1000 dt: 0.0025 default_connectivity_radius: 0.015 normalization_stats: [vel_mean, vel_std, acc_mean, acc_std] relations: required_datasets: [] optional_datasets: [] compatible_models: - id: OneScience/LagrangianMGN role: train_data required_for: [preflight, train, inference, evaluate, visualize] resource_ref: platform: modelscope repo_id: OneScience/LagrangianMGN repo_type: model url: https://modelscope.cn/models/OneScience/LagrangianMGN revision: main readme_path: README.md manifest_path: onescience_run_manifest.yaml expected_local_path: data/Water adapter: null run_matrix: scenarios: - name: dataset_validation default: true capability: preflight description: 验证 Water TFRecord 数据、metadata schema、文件清单和可读性。 required_datasets: - id: OneScience/lagrangian role: dataset_validation local_path: . required_model_files: [] required_dataset_files: [data/Water/metadata.json, data/Water/train.tfrecord, data/Water/valid.tfrecord, data/Water/test.tfrecord, files_sha256.jsonl] preconditions: - 已下载 OneScience/lagrangian 数据集仓库并 cd 到数据集仓库根目录。 command_refs: [commands.preflight.validate_dataset] expected_outputs: [dataset_validation_ok] capabilities: inference: false train: false finetune: false evaluate: false visualize: false deploy: false dataset_validation: true commands: download: - name: download_dataset target: dataset repo_id: OneScience/lagrangian local_path: /lagrangian required: true required_for: [dataset_validation, train, inference, evaluate] command: modelscope download --dataset OneScience/lagrangian preflight: - name: validate_dataset description: 验证 Water 数据集文件结构、metadata、TFRecord 首条样本和清单。 cwd: . command: python scripts/validate_lagrangian_dataset.py --dataset-root . required_files: [scripts/validate_lagrangian_dataset.py, data/Water/metadata.json, data/Water/train.tfrecord, data/Water/valid.tfrecord, data/Water/test.tfrecord, files_sha256.jsonl] required_env: [] expected_outputs: [{path: stdout, type: text, contains: "Lagrangian Water dataset validation passed"}] success_criteria: [exit_code == 0] prepare: [] inference: [] train: [] finetune: [] evaluate: [] visualize: [] deploy: [] expected_outputs: - id: dataset_validation_ok path: stdout type: text description_zh: 数据集验证输出 Lagrangian Water dataset validation passed。 diagnostics: - symptom: "missing TFRecord split" cause: 数据集下载不完整或当前目录不是数据集仓库根目录。 fix: 重新执行 modelscope download --dataset OneScience/lagrangian，并 cd 到下载后的数据集根目录。 - symptom: "metadata dim or sequence_length mismatch" cause: 使用了非 Water 或非本标准包的数据目录。 fix: 确认 ONESCIENCE_LAGRANGIAN_DATA_DIR 指向 OneScience/lagrangian 的 data/Water。 - symptom: "TensorFlow first-record check skipped" cause: 当前环境缺少 TensorFlow。 fix: 安装 TensorFlow 后重跑验证；文件大小和 SHA256 仍可先用于完整性判断。 domain_extension: cfd: dataset_family: DeepMind Learning to Simulate subset: Water simulation_type: lagrangian_particle_dynamics dimensionality: 2 splits: [train, valid, test]