37 lines
1.0 KiB
Python
37 lines
1.0 KiB
Python
import os
|
|
import pinecone
|
|
|
|
from database import *
|
|
import pandas as pd
|
|
|
|
from sentence_transformers import SentenceTransformer
|
|
|
|
from tqdm import tqdm
|
|
|
|
database_url = "sqlite:///jlm.db"
|
|
engine, Session = init_db_stuff(database_url)
|
|
|
|
PINECONE_KEY = os.getenv("PINECONE_API_DEFAULT")
|
|
pinecone.init(api_key=PINECONE_KEY, environment="us-west1-gcp")
|
|
index = pinecone.Index("movies")
|
|
|
|
model = SentenceTransformer("paraphrase-multilingual-MiniLM-L12-v2")
|
|
|
|
batch_size = 32
|
|
|
|
df = pd.read_sql("Select * from movies", engine)
|
|
df["combined_text"] = df["title"] + ": " + df["overview"].fillna('') + " - " + df["tagline"].fillna('') + " Genres:- " + df["genres"].fillna('')
|
|
|
|
print(f'Length of Combined Text: {len(df["combined_text"].tolist())}')
|
|
|
|
for x in tqdm(range(0,len(df),batch_size)):
|
|
to_send = []
|
|
trakt_ids = df["trakt_id"][x:x+batch_size].tolist()
|
|
sentences = df["combined_text"][x:x+batch_size].tolist()
|
|
embeddings = model.encode(sentences)
|
|
for idx, value in enumerate(trakt_ids):
|
|
to_send.append(
|
|
(
|
|
str(value), embeddings[idx].tolist()
|
|
))
|
|
index.upsert(to_send) |