OneScience-Group/MetNet-2

1{2  "model_name": "MetNet-2",3  "model_type": "metnet_2",4  "architectures": ["MetNet2"],5  "framework": "PyTorch",6  "domain": "weather",7  "task": "probabilistic-precipitation-forecasting",8  "implementation": {9    "entry_point": "model/metnet_2.py",10    "scope": "core-method and logical full-dimension sampled-window engineering reproduction",11    "train_script": "scripts/train.py",12    "inference_script": "scripts/inference.py",13    "evaluation_script": "scripts/result.py",14    "synthetic_data_script": "scripts/fake_data.py"15  },16  "architecture": {17    "logical_input_shape": ["B", 641, 512, 512],18    "logical_output_shape": ["B", 512, 512, 512],19    "engineering_window": [32, 32],20    "classes": 512,21    "lead_minutes": [2, 720, 2],22    "core": ["ConvLSTM", "lead-time FiLM", "dilated residual blocks", "spatial and class chunking"]23  },24  "data": {25    "datasets": ["MRMS", "HRRR", "GOES"],26    "format_version": "metnet2_selected_windows_v1",27    "input_channels": 641,28    "precipitation_range_mm_h": [0.0, 102.4],29    "coverage": "selected 32x32 target windows",30    "is_complete_global": false,31    "synthetic": true32  },33  "configuration_sources": [34    "conf/config.yaml",35    "model/metnet_2.py",36    "scripts/fake_data.py",37    "scripts/train.py",38    "scripts/inference.py",39    "scripts/result.py"40  ]41}42