Fix/bugs from langtest 230rc1

langtest/langtest.py

@@ -601,6 +601,7 @@ def generated_results(self) -> Optional[pd.DataFrame]:
         """
 
         column_order = [
+            "dataset_name",
             "model_name",
             "category",
             "test_type",
@@ -835,6 +836,7 @@ def testcases(self) -> pd.DataFrame:
                 testcases formatted into a pd.DataFrame
         """
         column_order = [
+            "dataset_name",
             "model_name",
             "category",
             "test_type",
@@ -889,6 +891,46 @@ def testcases(self) -> pd.DataFrame:
 
             testcases_df = pd.concat(testcases_df).reset_index(drop=True)
 
+        elif isinstance(self._testcases, dict) and isinstance(self.model, dict):
+            testcases_df = []
+            for k, v in self._testcases.items():
+                if isinstance(v, dict):
+                    model_testcases_df = pd.DataFrame.from_dict(
+                        [
+                            {
+                                "dataset_name": k,
+                                "model_name": model_name,
+                                **sample.to_dict(),
+                            }
+                            for model_name, samples in v.items()
+                            if isinstance(samples, list)
+                            for sample in samples
+                        ]
+                    )
+                else:
+                    model_testcases_df = pd.DataFrame.from_dict(
+                        [{"model_name": k, **x.to_dict()} for x in v]
+                    )
+
+                if "prompt" in model_testcases_df.columns:
+                    return model_testcases_df.fillna("-")
+
+                elif (
+                    "test_case" in model_testcases_df.columns
+                    and "original_question" in model_testcases_df.columns
+                ) and self.task != "political":
+                    model_testcases_df["original_question"].update(
+                        model_testcases_df.pop("test_case")
+                    )
+
+                testcases_df.append(model_testcases_df)
+            testcases_df = pd.concat(testcases_df).reset_index(drop=True)
+
+            columns = [c for c in column_order if c in testcases_df.columns]
+            testcases_df = testcases_df[columns]
+
+            return testcases_df.fillna("-")
+
         elif self.is_multi_dataset and isinstance(self._testcases, dict):
             testcases_df = pd.DataFrame(
                 [
@@ -1508,8 +1550,13 @@ def __multi_datasets_generate(self, dataset: Dict[str, list]):
                 print(f"{'':-^80}\n")
         elif isinstance(self.data, list) and self.__is_multi_model:
             temp_testcases = self.__single_dataset_generate(dataset)
-            for model_name, _ in self.model.items():
-                testcases[model_name] = [sample.copy() for sample in temp_testcases]
+            if isinstance(temp_testcases, dict) and set(temp_testcases.keys()) == set(
+                self.model.keys()
+            ):
+                testcases = temp_testcases
+            else:
+                for model_name, _ in self.model.items():
+                    testcases[model_name] = [sample.copy() for sample in temp_testcases]
 
         else:
             for dataset_name, samples in dataset.items():

langtest/transform/constants.py

@@ -614,7 +614,7 @@
                     "Taiwan", "Uruguay", "United States", "British Virgin Islands", "Virgin Islands"],
     "Low-income": ["Afghanistan", "Burundi", "Burkina Faso", "Central African Republic", "Congo", "Eritrea",
                    "Ethiopia", "Guinea", "Gambia", "Guinea-Bissau", "Liberia", "Madagascar", "Mali", "Mozambique",
-                   "Mtoalawi", "Niger", "North Korea", "Rwanda", "Sudan", "Sierra Leone", "Somalia",
+                   "Niger", "North Korea", "Rwanda", "Sudan", "Sierra Leone", "Somalia",
                    "South Sudan", "Syria", "Chad", "Togo", "Uganda", "Yemen", "Zambia"],
     "Lower-middle-income": ["Angola", "Benin", "Bangladesh", "Bolivia", "Bhutan", "Côte d'Ivoire", "Ivory Coast", "Cameroon",
                             "Congo", "Comoros", "Cabo Verde", "Djibouti", "Algeria", "Egypt",