save_dir = './case_result' config = { 'DEVICE': torch.device('cuda:0'), 'CAL_DEVICE': torch.device('cuda:2'), 'IN_LEN': 4, 'OUT_LEN': 1, 'BATCH_SIZE': 2, 'SCALE': 0.25, 'TASK': 'reg', 'DIM': 'HR', } data_loader = DataGenerator(data_path=global_config['DATA_PATH'], config=config) update_config(cfg, {'cfg': './params3.yaml'}) model = get_seg_model(cfg, 4, 1) model = torch.nn.DataParallel(model, device_ids=[0, 3]) model = model.to(config['DEVICE']) weight_path = '/home/warit/fcn/experiments/hrnet/model_logs/logs_4_1_03291123/model_28500.pth' model.load_state_dict(torch.load(weight_path, map_location='cuda')) files = sorted([file for file in glob.glob(global_config['DATA_PATH'])]) for i in tqdm(case): file_name = i[0] crop = i[1] sp = save_dir + '/' + file_name[:-4] if not os.path.exists(sp): os.makedirs(sp) test(model, data_loader, config, sp, files, file_name, crop=crop)
config=config), ]] data_loader = DataGenerator(data_path=global_config['DATA_PATH'], config=config) encoder = Encoder(convlstm_encoder_params[0], convlstm_encoder_params[1]).to(config['DEVICE']) forecaster = Forecaster(convlstm_forecaster_params[0], convlstm_forecaster_params[1], config=config).to(config['DEVICE']) encoder_forecaster = EF(encoder, forecaster).to(config['DEVICE']) weight_path = '/home/warit/senior/experiments/conv_logs/logs_4_10_2_False_05032140/model_25500.pth' encoder_forecaster.load_state_dict(torch.load(weight_path, map_location='cuda')) files = sorted([file for file in glob.glob(global_config['DATA_PATH'])]) for i in tqdm(case): file_name = i[0] crop = i[1] sp = save_dir + '/' + file_name[:-4] if not os.path.exists(sp): os.makedirs(sp) test(encoder_forecaster, data_loader, config, sp, files, file_name, crop=crop)