Merge branch 'main' into clem_homogeneize_generation_params

src/lighteval/metrics/metrics_sample.py

@@ -481,6 +481,7 @@ def __init__(
         self.normalize_gold = normalize_gold
         self.normalize_pred = normalize_pred
         self.tokenizer = tokenizer
+        self.scorer = None
 
     def compute(self, golds: list[str], predictions: list[str], **kwargs) -> float | dict:
         """Computes the metric(s) over a list of golds and predictions for one single sample.

src/lighteval/models/endpoints/endpoint_model.py

@@ -268,6 +268,7 @@ def __init__(  # noqa: C901
                     logger.error(
                         "Endpoint did not start within 30 minutes, there was a timeout. Please inspect the logs."
                     )
+                    self.cleanup()
                     raise e
                 except HfHubHTTPError as e:
                     # The endpoint actually already exists, we'll spin it up instead of trying to create a new one

src/lighteval/tasks/lighteval_task.py

@@ -130,7 +130,7 @@ def __post_init__(self):
         self.hf_avail_splits = tuple(self.hf_avail_splits) if self.hf_avail_splits is not None else None
         self.evaluation_splits = tuple(self.evaluation_splits)
         self.suite = tuple(self.suite)
-        self.stop_sequence = tuple(self.stop_sequence) if self.stop_sequence is not None else None
+        self.stop_sequence = tuple(self.stop_sequence) if self.stop_sequence is not None else ()
 
     def print(self):
         md_writer = MarkdownTableWriter()