The embedding is called twice

[Chen-Zile12]

AutoSchemaKG\atlas_rag\vectorstore\create_graph_index.py

def compute_text_embeddings(text_list, sentence_encoder: BaseEmbeddingModel, batch_size = 40, normalize_embeddings: bool = False):
"""Separated text embedding computation"""
text_embeddings = []

for i in tqdm(range(0, len(text_list), batch_size), desc="Encoding texts"):
    batch = text_list[i:i + batch_size]
    **embeddings = sentence_encoder.encode(batch, normalize_embeddings=normalize_embeddings)**
    if isinstance(embeddings, torch.Tensor):
        embeddings = embeddings.cpu().numpy()
    **text_embeddings.extend(sentence_encoder.encode(batch, normalize_embeddings = normalize_embeddings))**
return text_embeddings