tested on benchmark and small fixes to pyserini bm25 to spare

ieeta-pt · Oct 24, 2023 · 4e47275 · 4e47275
1 parent 06152f1
commit 4e47275
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 3 deletions.
diff --git a/converting_from_anserini.py b/converting_from_anserini.py
@@ -13,7 +13,7 @@
 def main(dataset_folder):
 
 
-    sparseCSR_collection = SparseCollectionCSR.from_bm25_pyserini_iterator(os.path.join(dataset_folder, "anserini_index"),
+    sparseCSR_collection = SparseCollectionCSR.from_bm25_pyserini_index(os.path.join(dataset_folder, "anserini_index"),
                                                                            k1=1.2,
                                                                            b=0.75,
                                                                            dtype=spare.float32,

diff --git a/spare/collection.py b/spare/collection.py
@@ -96,7 +96,7 @@ def from_bm25_pyserini_index(cls,
         sparse_collection = cls(index_reader.stats()["documents"], 
                                 vec_dim=index_reader.stats()["unique_terms"], 
                                 dtype=dtype, 
-                                weighting_schema=BM25WeightingSchema(k1=k1, b=b, k3=None, idf_weighting=idf_weighting),
+                                weighting_schema=BM25WeightingSchema(k1=k1, b=b, idf_weighting=None),
                                 backend=backend, 
                                 **kwargs)
 

diff --git a/spare/weighting_model.py b/spare/weighting_model.py
@@ -61,7 +61,7 @@ def _get_vars_to_save(self):
         return super()._get_vars_to_save() | {
             "k1": self.k1,
             "b": self.b,
-            "idf_weighting": idf_weighting
+            "idf_weighting": self.idf_weighting
             } 
 
     def _load_vars(self, data):