How can I make it require significantly less memory and start up faster?

[ErfolgreichCharismatisch]

This is how I use sentence-transformers on Windows 10 home

pkl = 'path/to/some.pkl'
query = 'some query'
from sentence_transformers import SentenceTransformer, util
model = SentenceTransformer('path/to/sentence-transformers_msmarco-distilbert-multilingual-en-de-v2-tmp-trained-scratch', device='cuda')

with open(pkl, "rb") as fIn:
  stored_data = pickle.load(fIn)
  doc_emb = stored_data['doc_emb']
  docs = stored_data['docs']


query_emb = model.encode(query.replace("\"",""), batch_size=6)
scores = util.dot_score(query_emb, doc_emb)[0].cpu().tolist()
del query_emb
//use results

The pkl is 1 GB in size, and when running the script, nothing else can run due to the amount of RAM and swapfile occupation.

The first run is also time consuming, then each query comes quickly.

How can I make it require significantly less memory and start up faster?