深層学習（Pytorch）を用いた、Kaggle Titanic実践 PART 10 （テストのCode）

予測するためのCode

import pickle
import pandas as pd
import numpy as np
import category_encoders as ce
import torch
from torch import nn
from data_preprocessing import data_preprocessing_test
from sklearn.model_selection import train_test_split
from sklearn import preprocessing

from model import Model
from earlystopping import EarlyStopping
from data_loader import *

model = Model()
model.load_state_dict(torch.load("models/model.pth"))
with open('models/ohe.pkl', 'rb') as f:
	ce_ohe = pickle.load(f)
with open('models/mm.pkl', 'rb') as f:
	mm = pickle.load(f)

model.eval()

data = pd.read_csv("titanic/test.csv")
X_test = data_preprocessing_test(data, ce_ohe)
X_test = mm.transform(X_test)

id_lst = list(data["PassengerId"].values)
pred_lst = list()
for data in X_test:
	res = model(torch.tensor(data).unsqueeze(0).float())
	pred_lst.extend(np.argmax(np.exp(res.detach().numpy()), axis=1))

sub_df = pd.DataFrame(np.array(id_lst+pred_lst).reshape(2, -1))
sub_df = sub_df.T
sub_df = sub_df.rename(columns={0: 'PassengerId', 1:"Survived"})
print(sub_df.head())
sub_df.to_csv("sub_dl.csv", index=False)