plotting was breaking in the case of cumulative and use_bos_token=Fal…

…se due to NaNs in `np.cumsum`; this commit provides workaround

surprisal/interface.py

@@ -17,7 +17,7 @@ def __init__(self, model_id=None) -> None:
         self.model_id = model_id
 
     @abstractmethod
-    def surprise(self, textbatch: typing.Union[typing.List, str]) -> "Surprisal":
+    def surprise(self, textbatch: typing.Union[typing.List, str]) -> "SurprisalArray":
         raise NotImplementedError
 
 
@@ -118,7 +118,13 @@ def lineplot(self, f=None, a=None, cumulative=False):
         if f is None or a is None:
             f, a = plt.subplots()
 
-        arr = np.cumsum(self.surprisals) if cumulative else self.surprisals
+        if cumulative:
+            arr = np.nan_to_num(self.surprisals.astype("float64"))
+            arr = np.cumsum(arr)
+            if arr[0] == 0:
+                arr[0] = float("nan")
+        else:
+            arr = self.surprisals
         a.plot(
             arr + np.random.rand(len(self)) / 10,
             ".--",