script.py

import pickle

import numpy as np
import pandas as pd
from sklearn.metrics.pairwise import cosine_similarity

test_df = pd.read_csv("/tmp/data/test.csv")

with open("model.pkl", "rb") as f:
    model = pickle.load(f)

scores = []
for _, row in test_df.iterrows():
    X_query = model["tokenizer"].transform([row["Query"]])
    is_cand = sum([(model["faq_ids"] == row[f"FAQ{i+1}"]).astype(int) for i in range(3)]) > 0
    sim = cosine_similarity(X_query, model["X_faq"][is_cand])[0]
    score = sim.max()
    scores.append(score)

predict = (np.array(scores) > model["thr"]).astype(int)

df = pd.DataFrame([(f"testid{i:04}", v) for i, v in enumerate(predict)], columns=["id", "pred"])
df.to_csv("submission.csv", index=None)