typo

miracl-reranking.py

@@ -36,7 +36,7 @@ def get_first_stage_runfile(lang):
         "bm25", "mdpr", "hybrid",
     ]
     return {
-        lang: f"https://huggingface.co/datasets/miracl/miracl-reranking/resolve/main/data/{first_stage}/{lang}.gz" for first_stage in first_stages
+        first_stage: f"https://huggingface.co/datasets/miracl/miracl-reranking/resolve/main/data/{first_stage}/{lang}.gz" for first_stage in first_stages
     }
 
 
@@ -78,7 +78,7 @@ def load_runfile(fn, topk=100):
     for line in file_handle:
         if not isinstance(line, str):
             line = line.decode()
-        qid, _, docid, _, score, _ = line.strip()
+        qid, _, docid, _, score, _ = line.strip().split()
         runs[qid][docid] = float(score)
 
     if topk > 0:
@@ -141,7 +141,6 @@ class MIRACLReranking(datasets.GeneratorBasedBuilder):
         miracl_corpus = datasets.load_dataset("miracl/miracl-corpus", lang)["train"]
         docid2doc = {doc["docid"]: formulate_doc(doc["title"], doc["text"]) for doc in miracl_corpus}
 
-        # topic_fn, qrel_fn = (filepaths) if len(filepaths) == 2 else (filepaths[0], None)
         topic_fn = filepaths["topics"]
         qrel_fn = filepaths["qrels"]
         runfile = filepaths["bm25"]
@@ -161,4 +160,3 @@ class MIRACLReranking(datasets.GeneratorBasedBuilder):
             data["candidates"] = [docid2doc[docid] for docid in runs[qid]]
 
             yield qid, data
-