data and retriever module

.gitattributes

@@ -52,3 +52,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.jpg filter=lfs diff=lfs merge=lfs -text
 *.jpeg filter=lfs diff=lfs merge=lfs -text
 *.webp filter=lfs diff=lfs merge=lfs -text
+*.idx filter=lfs diff=lfs merge=lfs -text

data/en/embs_IVF16384_HNSW32_2lvl_full.idx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4c8a9b6f6233a8927bc3981a9a84a86ab300cd4d6c9c381132fb9bcee82f3d4
+size 17268102277

data/en/paragraphs.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01430661566694a84fd77b7a9a9524525a53e87b1b6fe937f271f37a2e73678d
+size 7231592115

wiki_mpnet_index.py

@@ -0,0 +1,36 @@
+import faiss
+import json
+import numpy as np
+import zipfile
+
+from sentence_transformers import SentenceTransformer
+
+
+class Retriever:
+    def __init__(self):
+        self.index = faiss.read_index("data/en/embs_IVF16384_HNSW32_2lvl_full.idx")
+        self.index.nprobe = 128
+        self.model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2', device='cuda')
+        self.model.max_seq_length = 512
+        
+        
+    def get_paragraph_by_vec_idx(self, vec_idx):
+        archive = zipfile.ZipFile('data/en/paragraphs.zip', 'r')
+        chunk_id = vec_idx // 100000
+        line_id = vec_idx % 100000
+        with archive.open('enwiki_paragraphs_clean/enwiki_paragraphs_%03d.jsonl' % chunk_id) as f:
+            for i,l in enumerate(f):
+                if i == line_id:
+                    data = json.loads(l)
+                    break
+        return data
+        
+        
+    def search(self, query, k=5):
+        emb = self.model.encode(query)
+        _, neighbors = self.index.search(emb[np.newaxis, ...], k)
+        results = []
+        for n in neighbors[0]:
+            data = get_paragraph_by_vec_idx(n)
+            results.append(data)
+        return results