type hint improvements

syntaxgym.py

@@ -10,7 +10,8 @@ from copy import deepcopy
 import json
 from pathlib import Path
 import re
-from typing import List, Tuple
+from typing import List, Dict, Tuple
+from typing_extensions import TypedDict
 
 import datasets
 from datasets import logging
@@ -126,6 +127,11 @@ class SyntaxGym(datasets.GeneratorBasedBuilder):
             yield item["item_number"], item
 
 
+class SyntaxGymMetricResult(TypedDict):
+    prediction_results: List[List[bool]]
+    region_totals: List[Dict[Tuple[str, int], float]]
+
+
 class SyntaxGymMetric(datasets.Metric):
     """
     SyntaxGym prediction evaluation metric.
@@ -144,7 +150,8 @@ class SyntaxGymMetric(datasets.Metric):
         )
 
     def _compute(self, suite, model_id, batch_size: int = 16,
-                 add_start_token=True, device=None):
+                 add_start_token=True, device=None
+                 ) -> SyntaxGymMetricResult:
         if device is not None:
             assert device in ["gpu", "cpu", "cuda"]
             if device == "gpu":
@@ -258,7 +265,8 @@ class SyntaxGymMetric(datasets.Metric):
         return ret
 
     def compute_region_token_mapping(self, item, input_ids: torch.LongTensor,
-                                     offset_mapping: List[Tuple[int, int]]):
+                                     offset_mapping: List[Tuple[int, int]]
+                                     ) -> Dict[str, Dict[int, List[int]]]:
         # input_ids: B * T
         # offset_mapping: B * T * 2
         # assumes batch is sorted according to item's condition_name order