LLM/DocsGPT
1
0
Fork 0
mirror of https://github.com/arc53/DocsGPT.git synced 2025-11-29 08:33:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

raise more legible error if the embedding vector dims don't match

Browse Source
This commit is contained in:
asoderlind
2023-10-12 20:24:25 +02:00
parent e5e5a42736
commit 9b6ae46e92
1 changed files with 12 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
application/vectorstore/faiss.py
View File

@@ -2,6 +2,7 @@ from application.vectorstore.base import BaseVectorStore
from langchain.vectorstores import FAISS from langchain.vectorstores import FAISS
from application.core.settings import settings from application.core.settings import settings
HUGGINGFACE_MODEL_NAME = "huggingface_sentence-transformers/all-mpnet-base-v2"
class FaissStore(BaseVectorStore): class FaissStore(BaseVectorStore):
def __init__(self, path, embeddings_key, docs_init=None): def __init__(self, path, embeddings_key, docs_init=None):
@@ -12,10 +13,20 @@ class FaissStore(BaseVectorStore):
docs_init, self._get_embeddings(settings.EMBEDDINGS_NAME, embeddings_key) docs_init, self._get_embeddings(settings.EMBEDDINGS_NAME, embeddings_key)
) )
else: else:
embeddings = self._get_embeddings(settings.EMBEDDINGS_NAME, embeddings_key)
self.docsearch = FAISS.load_local( self.docsearch = FAISS.load_local(
self.path, self._get_embeddings(settings.EMBEDDINGS_NAME, settings.EMBEDDINGS_KEY) self.path, embeddings
) )
# Check that the word_embedding_dimension of the index matches the word_embedding_dimension of the embeddings
if settings.EMBEDDINGS_NAME == HUGGINGFACE_MODEL_NAME:
try:
word_embedding_dimension = embeddings.client[1].word_embedding_dimension
except AttributeError as e:
raise AttributeError("word_embedding_dimension not found in embeddings.client[1]") from e
if word_embedding_dimension != self.docsearch.index.d:
raise ValueError("word_embedding_dimension != docsearch_index_word_embedding_dimension")
def search(self, *args, **kwargs): def search(self, *args, **kwargs):
return self.docsearch.similarity_search(*args, **kwargs) return self.docsearch.similarity_search(*args, **kwargs)