为所有masked的state预测评估函数添加mask_val，修复遗留问题

libcity/config/model/traffic_state_pred/STSSL.json

@@ -34,5 +34,5 @@
   "pad_back_trend": 0,
   "interval_period": 1,
   "interval_trend": 7,
-  "eval_mask_val": 5.001
+  "mask_val": 5.001
 }

libcity/evaluator/traffic_state_evaluator.py

@@ -13,10 +13,10 @@ class TrafficStateEvaluator(AbstractEvaluator):
     def __init__(self, config):
         self.metrics = config.get('metrics', ['MAE'])  # 评估指标, 是一个 list
         self.allowed_metrics = ['MAE', 'MSE', 'RMSE', 'MAPE', 'masked_MAE', 'masked_MSE', 'masked_RMSE', 'masked_MAPE',
-                                'R2', 'EVAR', "IN_masked_MAE", "IN_masked_MAPE", "OUT_masked_MAE", "OUT_masked_MAPE"]
+                                'R2', 'EVAR']
         self.save_modes = config.get('save_mode', ['csv', 'json'])
         self.mode = config.get('evaluator_mode', 'single')  # or average
-        self.mask_val = config.get('eval_mask_val', None)
+        self.mask_val = config.get('mask_val', None)
         self.config = config
         self.len_timeslots = 0
         self.result = {}  # 每一种指标的结果
@@ -60,10 +60,12 @@ def collect(self, batch):
                                                   mask_val=self.mask_val).item())
                     elif metric == 'masked_MSE':
                         self.intermediate_result[metric + '@' + str(i)].append(
-                            loss.masked_mse_torch(y_pred[:, :i], y_true[:, :i], 0).item())
+                            loss.masked_mse_torch(y_pred[:, :i], y_true[:, :i], 0,
+                                                  mask_val=self.mask_val).item())
                     elif metric == 'masked_RMSE':
                         self.intermediate_result[metric + '@' + str(i)].append(
-                            loss.masked_rmse_torch(y_pred[:, :i], y_true[:, :i], 0).item())
+                            loss.masked_rmse_torch(y_pred[:, :i], y_true[:, :i], 0,
+                                                  mask_val=self.mask_val).item())
                     elif metric == 'masked_MAPE':
                         self.intermediate_result[metric + '@' + str(i)].append(
                             loss.masked_mape_torch(y_pred[:, :i], y_true[:, :i], 0,
@@ -95,10 +97,12 @@ def collect(self, batch):
                                                   mask_val=self.mask_val).item())
                     elif metric == 'masked_MSE':
                         self.intermediate_result[metric + '@' + str(i)].append(
-                            loss.masked_mse_torch(y_pred[:, i - 1], y_true[:, i - 1], 0).item())
+                            loss.masked_mse_torch(y_pred[:, i - 1], y_true[:, i - 1], 0,
+                                                  mask_val=self.mask_val).item())
                     elif metric == 'masked_RMSE':
                         self.intermediate_result[metric + '@' + str(i)].append(
-                            loss.masked_rmse_torch(y_pred[:, i - 1], y_true[:, i - 1], 0).item())
+                            loss.masked_rmse_torch(y_pred[:, i - 1], y_true[:, i - 1], 0,
+                                                  mask_val=self.mask_val).item())
                     elif metric == 'masked_MAPE':
                         self.intermediate_result[metric + '@' + str(i)].append(
                             loss.masked_mape_torch(y_pred[:, i - 1], y_true[:, i - 1], 0,

libcity/model/loss.py

@@ -56,7 +56,7 @@ def quantile_loss(preds, labels, delta=0.25):
     return torch.mean(torch.where(condition, large_res, small_res))
 
 
-def masked_mape_torch(preds, labels, null_val=np.nan, eps=0, mask_val=np.nan):
+def masked_mape_torch(preds, labels, null_val=np.nan, eps=0, mask_val=None):
     labels[torch.abs(labels) < 1e-4] = 0
     if np.isnan(null_val) and eps != 0:
         loss = torch.abs((preds - labels) / (labels + eps))
@@ -65,7 +65,7 @@ def masked_mape_torch(preds, labels, null_val=np.nan, eps=0, mask_val=np.nan):
         mask = ~torch.isnan(labels)
     else:
         mask = labels.ne(null_val)
-    if not np.isnan(mask_val):
+    if mask_val:
         mask &= labels.ge(mask_val)
     mask = mask.float()
     mask /= torch.mean(mask)
@@ -76,12 +76,14 @@ def masked_mape_torch(preds, labels, null_val=np.nan, eps=0, mask_val=np.nan):
     return torch.mean(loss)
 
 
-def masked_mse_torch(preds, labels, null_val=np.nan):
+def masked_mse_torch(preds, labels, null_val=np.nan, mask_val=None):
     labels[torch.abs(labels) < 1e-4] = 0
     if np.isnan(null_val):
         mask = ~torch.isnan(labels)
     else:
         mask = labels.ne(null_val)
+    if mask_val:
+        mask &= labels.ge(mask_val)
     mask = mask.float()
     mask /= torch.mean(mask)
     mask = torch.where(torch.isnan(mask), torch.zeros_like(mask), mask)
@@ -91,10 +93,10 @@ def masked_mse_torch(preds, labels, null_val=np.nan):
     return torch.mean(loss)
 
 
-def masked_rmse_torch(preds, labels, null_val=np.nan):
+def masked_rmse_torch(preds, labels, null_val=np.nan, mask_val=None):
     labels[torch.abs(labels) < 1e-4] = 0
     return torch.sqrt(masked_mse_torch(preds=preds, labels=labels,
-                                       null_val=null_val))
+                                       null_val=null_val, mask_val=mask_val))
 
 
 def r2_score_torch(preds, labels):