Fix typo in dense retrieval doc (#327)

castorini · Jan 17, 2021 · 889a976 · 889a976
1 parent a8c9443
commit 889a976
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 3 deletions.
diff --git a/docs/dense-retrieval.md b/docs/dense-retrieval.md
@@ -41,7 +41,8 @@ recall_1000             all     0.9618
 ```
 
 > To evaluate with on-the-fly query encoding, replace `--encoded-queries` with our pretrained encoder model
-> `--encoder tct_colbert-msmarco`. The encoding will run on CPU by default. To enable GPU, add `--device cuda:0`
+> on [Hugging Face](https://huggingface.co/castorini/tct_colbert-msmarco/tree/main) 
+> `--encoder castorini/tct_colbert-msmarco`. The encoding will run on CPU by default. To enable GPU, add `--device cuda:0`.
 > NOTE: Using GPU query encoding will give slightly different result. (E.g. MRR @10: 0.3349694137444839)
 
 
@@ -82,7 +83,8 @@ recall_1000             all     0.9640
 
 You'll notice that hnsw index leads to a small loss in effectiveness.
 > To evaluate with on-the-fly query encoding, replace `--encoded-queries` with our pretrained encoder model
-> `--encoder tct_colbert-msmarco`. The encoding will run on CPU by default. To enable GPU, add `--device cuda:0`
+> on [Hugging Face](https://huggingface.co/castorini/tct_colbert-msmarco/tree/main) 
+> `--encoder castorini/tct_colbert-msmarco`. The encoding will run on CPU by default. To enable GPU, add `--device cuda:0`.
 > NOTE: Using GPU query encoding will give slightly different result. (E.g. MRR @10: 0.3349479237731372)
 
 

diff --git a/scripts/msmarco-passage/build_hnsw_index.py b/scripts/msmarco-passage/build_hnsw_index.py
@@ -24,4 +24,4 @@
     print(vectors)
     print('Indexing')
     hnsw_index.add(vectors)
-    faiss.write_index(hnsw_index, os.path.join(args.bf_index, 'index'))
+    faiss.write_index(hnsw_index, os.path.join(args.hnsw_index, 'index'))