src/training/model_configs/RN50.json

{
    "embed_dim": 512,
    "vision_block": "bottleneck",
    "input_channels": 5,
    "vision_layers": [
        3,
        4,
        6,
        3
    ],
    "vision_width": 64,
    "input_size": 1024,
    "molecule_layers": 4,
    "hidden_dim": 1024
}