{
  "model_name": "Streamflow-LSTM",
  "model_type": "gauge_specific_stacked_lstm",
  "architectures": ["StreamflowLSTM"],
  "framework": "PyTorch",
  "torch_dtype": "float32",
  "transformers_version": "4.44.0",
  "domain": "hydrology",
  "task": "six-hourly-streamflow-forecasting",
  "license": "apache-2.0",
  "paper": {
    "title": "Using a long short-term memory (LSTM) neural network to boost river streamflow forecasts over the western United States",
    "doi": "10.5194/hess-26-5449-2022"
  },
  "implementation": {
    "entry_point": "model/streamflow_lstm.py",
    "scope": "core-method reduced-ensemble engineering reproduction",
    "train_script": "scripts/train.py",
    "inference_script": "scripts/inference.py",
    "evaluation_script": "scripts/result.py",
    "synthetic_data_script": "scripts/fake_data.py"
  },
  "architecture": {
    "input_shape": ["B", 28, 23],
    "output_shape": ["B"],
    "paper_hidden_sizes": [50, 50, 50],
    "activations": ["ReLU", "ReLU", "tanh"],
    "dense_outputs": 1
  },
  "data": {
    "format_version": "streamflow-lstm-v1",
    "gauges": 10,
    "input_shape": ["B", 28, 23],
    "forecast_steps": 40,
    "step_hours": 6,
    "unit": "m3 s-1",
    "synthetic": true
  },
  "checkpoint": {
    "path": "result/checkpoints/streamflow_lstm.pt",
    "format": "single-file multi-gauge multi-member PyTorch checkpoint",
    "official_weights": false
  },
  "configuration_sources": [
    "conf/config.yaml",
    "model/streamflow_lstm.py",
    "scripts/fake_data.py",
    "scripts/train.py",
    "scripts/inference.py",
    "scripts/result.py"
  ]
}