Add initial prototype for scored-docs access into ir_datasets #1

data/ir_datasets_scored_docs.py

@@ -0,0 +1,15 @@
+#!/usr/bin/env python3
+from ir_datasets.util import RequestsDownload, Cache, home_path
+from ir_datasets.formats import TrecScoredDocs
+
+
+def scored_docs(rank_distill_llm_run='__rankzephyr-colbert-10000-sampled-100__msmarco-passage-train-judged.run'):
+    base_path = home_path() / 'rank-disti-llm'
+    requests_download = RequestsDownload(f'https://zenodo.org/records/11147862/files/{rank_distill_llm_run}?download=1')
+    scored_docs = TrecScoredDocs(Cache(requests_download, base_path/rank_distill_llm_run))
+
+    return scored_docs
+
+if __name__ == '__main__':
+    for i in scored_docs().scoreddocs_iter():
+        print(i)