| from main import * | |
| def get_satclip(ckpt_path, device, return_all=False): | |
| ckpt = torch.load(ckpt_path,map_location=device) | |
| ckpt['hyper_parameters'].pop('eval_downstream') | |
| ckpt['hyper_parameters'].pop('air_temp_data_path') | |
| ckpt['hyper_parameters'].pop('election_data_path') | |
| lightning_model = SatCLIPLightningModule(**ckpt['hyper_parameters']).to(device) | |
| lightning_model.load_state_dict(ckpt['state_dict']) | |
| lightning_model.eval() | |
| geo_model = lightning_model.model | |
| if return_all: | |
| return geo_model | |
| else: | |
| return geo_model.location |