basic: dataset: SolarEnergy device: cuda:0 mode: train model: AGCRN seed: 2023 data: batch_size: 64 column_wise: false days_per_week: 7 horizon: 24 input_dim: 1 lag: 24 normalizer: std num_nodes: 137 steps_per_day: 24 test_ratio: 0.2 val_ratio: 0.2 model: batch_size: 64 dropout: 0.3 gcn_depth: 1 gcn_num: 2 hid_c: 64 in_dim: 1 num_nodes: 137 out_dim: 1 residual_channels: 32 skip_channels: 64 subgraph_size: 20 tanhalpha: 3 time_strides: 1 train: batch_size: 64 debug: false early_stop: true early_stop_patience: 15 epochs: 100 grad_norm: false log_step: 100 loss_func: mae lr_decay: true lr_decay_rate: 0.3 lr_decay_step: 5,20,40,70 lr_init: 0.003 mae_thresh: mape_thresh: 0.001 max_grad_norm: 5 output_dim: 1 plot: false real_value: true seed: 10 weight_decay: 0