[Python] Upserting in Pinecone takes too long

I'm trying to upsert reviews that i've scraped into pinecone. For the embedding model im using jina-embedding-v3. For 204 reviews this takes around 2.5 hours! in Colab. Tried using GPU but the embeddings arent using GPU. Am i doing something wrong? Is there a way that i can speed up the process? The code is below:

Initialising DB:

if index_name not in pc.list_indexes().names():
pc.create_index(
name=index_name,
dimension=1024,
metric="cosine",
spec=ServerlessSpec(
cloud="aws",
region="us-east-1"
)
)


Embedding & Upserting:

device = 'cuda' if torch.cuda.is_available() else 'cpu'

# Load the Jina embedding model and tokenizer from Hugging Face
model_name = "jinaai/jina-embeddings-v3"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModel.from_pretrained(model_name, trust_remote_code=True)

from langchain.text_splitter import SpacyTextSplitter
text_splitter = SpacyTextSplitter(chunk_size=500)

# Function to generate embeddings
def generate_embeddings(text, task='retrieval.passage'):
return model.encode(text, convert_to_tensor=True, task=task).numpy()

for review_id, review in enumerate(all_reviews[:2]):
chunks = text_splitter.split_text(review)

for chunk_index, chunk in enumerate(chunks):
embedding = generate_embeddings(chunk)

unique_id = f"{review_id}_{chunk_index}"

metadata = {"review_id": review_id, "chunk_index": chunk_index, "text": chunk}

index.upsert([(unique_id, embedding, metadata)])

# Generate and store embeddings
for review_id, review in enumerate(all_reviews):
chunks = text_splitter.split_text(review)

for chunk_index, chunk in enumerate(chunks):
embedding = generate_embeddings(chunk)

unique_id = f"{review_id}_{chunk_index}"

metadata = {"review_id": review_id, "chunk_index": chunk_index, "text": chunk}

index.upsert([(unique_id, embedding, metadata)])

Continue reading...