better model for the retriever: msmarco-distilbert-base-v4
Browse files
app.py
CHANGED
@@ -41,7 +41,7 @@ model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float32
|
|
41 |
|
42 |
# sentence transformers to be used in vector store
|
43 |
embeddings = HuggingFaceEmbeddings(
|
44 |
-
model_name="sentence-transformers/
|
45 |
model_kwargs={'device': 'cpu'},
|
46 |
encode_kwargs={'normalize_embeddings': False}
|
47 |
)
|
@@ -63,7 +63,7 @@ def prepare_vector_store_retriever(filename):
|
|
63 |
documents = text_splitter.split_documents(raw_documents)
|
64 |
|
65 |
# Creating a vectorstore
|
66 |
-
vectorstore = FAISS.from_documents(documents, embeddings, distance_strategy=DistanceStrategy.
|
67 |
|
68 |
return VectorStoreRetriever(vectorstore=vectorstore, search_kwargs={"k": 2})
|
69 |
|
|
|
41 |
|
42 |
# sentence transformers to be used in vector store
|
43 |
embeddings = HuggingFaceEmbeddings(
|
44 |
+
model_name="sentence-transformers/msmarco-distilbert-base-v4",
|
45 |
model_kwargs={'device': 'cpu'},
|
46 |
encode_kwargs={'normalize_embeddings': False}
|
47 |
)
|
|
|
63 |
documents = text_splitter.split_documents(raw_documents)
|
64 |
|
65 |
# Creating a vectorstore
|
66 |
+
vectorstore = FAISS.from_documents(documents, embeddings, distance_strategy=DistanceStrategy.DOT_PRODUCT)
|
67 |
|
68 |
return VectorStoreRetriever(vectorstore=vectorstore, search_kwargs={"k": 2})
|
69 |
|