type ignores

src/gluonts/itertools.py

@@ -298,7 +298,7 @@ def split_into(xs: Sequence, n: int) -> Sequence:
     # e.g. 10 by 3 -> 4, 3, 3
     relative_splits[:remainder] += 1
 
-    return split(xs, np.cumsum(relative_splits))
+    return split(xs, np.cumsum(relative_splits))  # type: ignore[arg-type]
 
 
 @dataclass

src/gluonts/model/forecast.py

@@ -178,14 +178,14 @@ class Quantile:
     @classmethod
     def from_float(cls, quantile: float) -> "Quantile":
         assert isinstance(quantile, float)
-        return cls(value=quantile, name=str(quantile))
+        return cls(value=quantile, name=str(quantile))  # type: ignore[call-arg]
 
     @classmethod
     def from_str(cls, quantile: str) -> "Quantile":
         assert isinstance(quantile, str)
 
         try:
-            return cls(value=float(quantile), name=quantile)
+            return cls(value=float(quantile), name=quantile)  # type: ignore[call-arg]
         except ValueError:
             m = re.match(r"^p(\d+)$", quantile)
 

src/gluonts/mx/trainer/learning_rate_scheduler.py

@@ -247,8 +247,8 @@ def __init__(
             0 <= min_lr <= base_lr
         ), "The value of `min_lr` should be >= 0 and <= base_lr"
 
-        self.lr_scheduler = MetricAttentiveScheduler(
-            patience=Patience(
+        self.lr_scheduler = MetricAttentiveScheduler(  # type: ignore[call-arg]
+            patience=Patience(  # type: ignore[call-arg]
                 patience=patience, objective=Objective.from_str(objective)
             ),
             learning_rate=base_lr,

src/gluonts/torch/model/patch_tst/module.py

@@ -229,7 +229,7 @@ def forward(
             # shift time features by `prediction_length` so that they are
             # aligned with the target input.
             time_feat = take_last(
-                torch.cat((past_time_feat, future_time_feat), dim=1),
+                torch.cat((past_time_feat, future_time_feat), dim=1),  # type: ignore[arg-type]
                 dim=1,
                 num=self.context_length,
             )

src/gluonts/torch/model/tft/layers.py

@@ -27,7 +27,7 @@ def forward(self, features: torch.Tensor) -> List[torch.Tensor]:  # type: ignore
         concat_features = super().forward(features=features)
 
         if self._num_features > 1:
-            return torch.chunk(concat_features, self._num_features, dim=-1)
+            return torch.chunk(concat_features, self._num_features, dim=-1)  # type: ignore[return-value]
         else:
             return [concat_features]
 

src/gluonts/torch/modules/feature.py

@@ -38,7 +38,7 @@ def forward(self, features: torch.Tensor) -> torch.Tensor:
                 features, self._num_features, dim=-1
             )
         else:
-            cat_feature_slices = [features]
+            cat_feature_slices = [features]  # type: ignore[assignment]
 
         return torch.cat(
             [

src/gluonts/zebras/_time_frame.py

@@ -453,11 +453,11 @@ def split(
 
         # If past_length is not provided, it will equal to `index`, since
         # `len(tf.split(5).past) == 5`
-        past_length: int = maybe.unwrap_or(past_length, index)
+        past_length: int = maybe.unwrap_or(past_length, index)  # type: ignore[annotation-unchecked]
 
         # Same logic applies to future_length, except that we deduct from the
         # right. (We can't use past_length, since it can be unequal to index).
-        future_length: int = maybe.unwrap_or(future_length, len(self) - index)
+        future_length: int = maybe.unwrap_or(future_length, len(self) - index)  # type: ignore[annotation-unchecked]
 
         if self.index is None:
             new_index = None