Upload tasktemplates.py

tasktemplates.py

@@ -85,20 +85,36 @@ class TaskTemplates(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        return [
-            datasets.SplitGenerator(
-                name=datasets.Split.TRAIN,
-                gen_kwargs={"data_url": self.config.data_url}
+        urls_to_download = {
+            'train': []
+        }
+        for dataset in _DATASETS:
+            urls_to_download['train'].append(
+                f"{_PREFIX}{dataset}/templates.yaml")
+
+        downloaded_files = dl_manager.download_and_extract(urls_to_download)
+        splits = []
+
+        if "train" in downloaded_files:
+            splits.append(
+                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={
+                                        'filepath': downloaded_files["train"]})
             )
-        ]
 
-    def _generate_examples(self, data_url):
-        with open(data_url, 'r') as f:
-            data = yaml.safe_load(f)
+        return splits
 
-        for task, templates in data.items():
-            for template in templates:
-                yield task, {
-                    "task": task,
-                    "templates": template
-                }
+    def _generate_examples(self, filepath):
+        id = 0
+        for path in filepath:
+            with open(path, 'r') as file:
+                template = yaml.load(file, Loader=yaml.FullLoader)
+                for _, task_templates in template['templates'].items():
+                    for task_template in task_templates:
+                        yield id, {
+                            "name": task_template['name'],
+                            "input": task_template['input'],
+                            "target": task_template['target'],
+                            "languages": task_template['metadata']['languages'],
+                            "metrics": task_template['metadata']['metrics']
+                        }
+                        id += 1