Fix bugs

FlagEmbedding/evaluation/__init__.py

@@ -1,7 +0,0 @@
-from .air_bench import AIRBenchEvalModelArgs, AIRBenchEvalArgs, AIRBenchEvalRunner
-from .beir import *
-# from miracle import *
-# from mkqa import *
-# from mldr import *
-# from msmarco import *
-from mteb import *

FlagEmbedding/evaluation/beir/data_loader.py

@@ -409,7 +409,7 @@ def _load_local_qrels(self, save_dir: str, dataset_name: Optional[str] = None, s
         Returns:
             datasets.DatasetDict: A dict of relevance of query and document.
         """
-        checked_split = self.check_splits(split)
+        checked_split = self.check_splits(split, dataset_name=dataset_name)
         if len(checked_split) == 0:
             raise ValueError(f"Split {split} not found in the dataset.")
         split = checked_split[0]
@@ -450,7 +450,7 @@ def _load_local_queries(self, save_dir: str, dataset_name: Optional[str] = None,
         Returns:
             datasets.DatasetDict: A dict of queries with id as key, query text as value.
         """
-        checked_split = self.check_splits(split)
+        checked_split = self.check_splits(split, dataset_name=dataset_name)
         if len(checked_split) == 0:
             raise ValueError(f"Split {split} not found in the dataset.")
         split = checked_split[0]

scripts/hn_mine.py

@@ -159,6 +159,12 @@ def find_knn_neg(
     p_vecs = model.encode(corpus)
     print(f'inferencing embedding for queries (number={len(queries)})--------------')
     q_vecs = model.encode_queries(queries)
+
+    # check if the embeddings are in dictionary format: M3Embedder
+    if isinstance(p_vecs, dict):
+        p_vecs = p_vecs["dense_vecs"]
+    if isinstance(q_vecs, dict):
+        q_vecs = q_vecs["dense_vecs"]
 
     print('create index and search------------------')
     index = create_index(p_vecs, use_gpu=use_gpu)