model:

    start_epoch: 0
    max_epoch: 10
    lr: 5e-4
    beta_1: 0.9
    beta_2: 0.99
    weight_decay: 1e-5
    warmup_epochs: 10000
    warmup_start_lr: 1e-8
    eta_min: 1e-8

    # ClimaX architecture params (from official config)
    img_size: [32, 64]               # 5.625° resolution (lat, lon)
    patch_size: 2
    embed_dim: 1024
    depth: 8
    decoder_depth: 2
    num_heads: 16
    mlp_ratio: 4.0
    drop_path: 0.1
    drop_rate: 0.1

    # Forecasting settings
    # NOTE: ERA5 data has 6-hour time steps (time_step=6 in HDF5 files).
    # The lead_time passed to the model is (predict_range * hrs_each_step) / 100.
    # For 6-hour forecast: predict_range=6, hrs_each_step=1 → lead_time=0.06
    # For 72-hour forecast: predict_range=72, hrs_each_step=1 → lead_time=0.72
    # The model conditions on lead_time; it must match the actual data gap.
    predict_range: 6                  # forecast lead time in hours
    hrs_each_step: 1                  # hours per data step factor for lead_time computation

    checkpoint_dir: "./data/checkpoints"
    patience: 50

# 整个数据读取流程
datapipe:
  name: "ERA5"
  task: "weather_forecasting"

  # dataset设定
  dataset:
    type: "hdf5"
    data_dir: './data/' # "$ONESCIENCE_DATASETS_DIR/ERA5/newh5/"

    train_time: [2000, 2001]
    val_time: [2002]
    test_time: [2003]
    img_size: [32, 64]
    verbose: true
    cache: false

    # ClimaX 48 input variables: 3 constants + 3 surface + 42 pressure-level
    channels:
      # Static constants (3)
      - "land_sea_mask"
      - "orography"
      - "lattitude"

      # Surface variables (3)
      - "2m_temperature"
      - "10m_u_component_of_wind"
      - "10m_v_component_of_wind"

      # Geopotential at 7 pressure levels
      - "geopotential_50"
      - "geopotential_250"
      - "geopotential_500"
      - "geopotential_600"
      - "geopotential_700"
      - "geopotential_850"
      - "geopotential_925"

      # U component of wind at 7 pressure levels
      - "u_component_of_wind_50"
      - "u_component_of_wind_250"
      - "u_component_of_wind_500"
      - "u_component_of_wind_600"
      - "u_component_of_wind_700"
      - "u_component_of_wind_850"
      - "u_component_of_wind_925"

      # V component of wind at 7 pressure levels
      - "v_component_of_wind_50"
      - "v_component_of_wind_250"
      - "v_component_of_wind_500"
      - "v_component_of_wind_600"
      - "v_component_of_wind_700"
      - "v_component_of_wind_850"
      - "v_component_of_wind_925"

      # Temperature at 7 pressure levels
      - "temperature_50"
      - "temperature_250"
      - "temperature_500"
      - "temperature_600"
      - "temperature_700"
      - "temperature_850"
      - "temperature_925"

      # Relative humidity at 7 pressure levels
      - "relative_humidity_50"
      - "relative_humidity_250"
      - "relative_humidity_500"
      - "relative_humidity_600"
      - "relative_humidity_700"
      - "relative_humidity_850"
      - "relative_humidity_925"

      # Specific humidity at 7 pressure levels
      - "specific_humidity_50"
      - "specific_humidity_250"
      - "specific_humidity_500"
      - "specific_humidity_600"
      - "specific_humidity_700"
      - "specific_humidity_850"
      - "specific_humidity_925"

    # ClimaX output variables (5)
    out_variables:
      - "geopotential_500"
      - "temperature_850"
      - "2m_temperature"
      - "10m_u_component_of_wind"
      - "10m_v_component_of_wind"

    # Short names for evaluation
    variables:
      - "z500"
      - "t850"
      - "t2m"
      - "u10"
      - "v10"

    # 时间配置
    time_range: ["2000-01-01", "2020-12-31"]
    time_steps: 1
    time_res: 6

    # 空间配置
    spatial_resolution: [5.625, 5.625]

    # 采样配置
    num_samples: -1  # -1 表示使用全部数据
    shuffle: true
    random_seed: 42

    # 领域特定配置
    extra:
      levels: [500, 850]
      lat_range: [-90, 90]
      lon_range: [0, 360]

  # 数据转换配置
  transforms:
    - type: "Normalize"
      params:
        mean: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
        std: [1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
        keys: ["input", "target"]

    - type: "ToTensor"
      params:
        keys: null  # null表示转换所有numpy数组

  # DataLoader配置
  dataloader:
    mask_dtype: "float32"
    batch_size: 1
    num_workers: 1
    pin_memory: true
    drop_last: true
    shuffle: false  # 使用sampler时设为false
    prefetch_factor: 2
    persistent_workers: true

  # 分布式配置
  distributed:
    enabled: true
    sampler: "DistributedSampler"
    rank: 0
    world_size: 4
    shuffle: true
    seed: 42
    drop_last: true