fix: use time_intervals to calculate STPGCN points_per_hour parameter (…

…LibCity#425) * fix: use time_intervals to calculate points_per_hour * fix: remove time_intervals parameter
Lucas-lyh · Jul 2, 2024 · a0c81ca · a0c81ca
1 parent 04232be
commit a0c81ca
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 4 deletions.
diff --git a/libcity/config/data/STPGCNDataset.json b/libcity/config/data/STPGCNDataset.json
@@ -11,7 +11,6 @@
   "input_window": 12,
   "output_window": 12,
 
-  "points_per_hour": 12,
   "alpha": 4,
   "beta": 2
 }
diff --git a/libcity/data/dataset/dataset_subclass/stpgcn_dataset.py b/libcity/data/dataset/dataset_subclass/stpgcn_dataset.py
@@ -38,7 +38,7 @@ def __init__(self, config):
                                             'point_based_{}.npz'.format(self.parameters_str))
 
         self.feature_name = {'X': 'float', 'y': 'float', 'pos_w': 'int', 'pos_d': 'int'}
-        self.points_per_hour = config.get('points_per_hour', 12)
+        self.points_per_hour = self.time_intervals // 60
         self.alpha = config.get('alpha', 4)
         self.beta = config.get('beta', 2)
         self.t_size = self.beta + 1
@@ -60,7 +60,7 @@ def get_data_feature(self):
         """
         return {"scaler": self.scaler, "ext_dim": self.ext_dim, "spatial_distance": self.spatial_distance,
                 "range_mask": self.range_mask, "num_nodes": self.num_nodes, "feature_dim": self.feature_dim,
-                "output_dim": self.output_dim, "num_batches": self.num_batches}
+                "output_dim": self.output_dim, "num_batches": self.num_batches, "points_per_hour": self.points_per_hour}
 
     def _load_cache_train_val_test(self):
         self._logger.info('Loading ' + self.cache_file_name)

diff --git a/libcity/model/traffic_flow_prediction/STPGCN.py b/libcity/model/traffic_flow_prediction/STPGCN.py
@@ -381,7 +381,7 @@ def __init__(self, config, data_feature):
         self.beta = config.get("beta", 2)
         self.t_size = self.beta + 1
         self.week_len = 7
-        self.day_len = config.get("points_per_hour") * 24
+        self.day_len = self.data_feature.get("points_per_hour") * 24
         self.range_mask = torch.Tensor(self.range_mask).to(self.device)
 
         self.PAD = GeneratePad(self.device, self.C, self.V, self.d, self.beta)