def run():
# Training settings
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser.add_argument('--test-batch-size', type=int, default=1000, metavar='N',
help='input batch size for testing (default: 1000)')
parser.add_argument('--no-cuda', action='store_true', default=True,
help='disables CUDA training')
parser.add_argument('--seed', type=int, default=1, metavar='S',
help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=10, metavar='N',
help='how many batches to wait before logging training status')
parser.add_argument('--test-AWGN', action='store_true', default=False,
help='Test AWGN transformation')
parser.add_argument('--sigma', type=float, default=0.0, metavar='N',
help='standard deviation of AWGN')
args = parser.parse_args()
use_cuda = not args.no_cuda and torch.cuda.is_available()
torch.manual_seed(args.seed)
device = torch.device("cuda" if use_cuda else "cpu")
test_kwargs = {'batch_size': args.test_batch_size}
## import model from main and import weights
model = main.Net().to(device)
model.load_state_dict(torch.load("mnist_cnn.pt"))
# Print model's state_dict
print("Model's state_dict:")
for param_tensor in model.state_dict():
print(param_tensor, "\t", model.state_dict()[param_tensor].size())
for i, sigma in enumerate([0, 0.3, 0.6, 1.0]):
transform = transforms.Compose([transforms.ToTensor(), main.AWGN(sigma), transforms.Normalize((0.1307,), (0.3081,))])
test_dataset = datasets.MNIST('../data', train=False, transform=transform)
test_loader = torch.utils.data.DataLoader(test_dataset, **test_kwargs)
print(f"\nAccuracy for AWGN with a standard deviation of {sigma}")
main.test(model, device, test_loader)
actions[16] = (generate_actions(16, list(np.arange(1001, 1500, 99) / 1000), df,
False)) # PRIMARY MORTGAGE INSURANCE PERCENT
actions[18] = (generate_actions(18, list(np.arange(1001, 1500, 99) / 1000), df,
False)) # CoBorrower Credit Score
actions[19] = (generate_actions(19, list(range(1, 4)), df,
True)) #MORTGAGE INSURANCE TYPE
actions[20] = (generate_actions(20, list(range(1, 3)), df,
True)) #RELOCATION MORTGAGE INDICATOR
# Flattened list of actions
actions_list = []
for key in actions:
actions_list.extend(actions[key])
MEAN_ACTION_COST = statistics.mean(
[action.cost for action in actions_list]
) # List Comprehension - create a list of only the costs of all of the actions
features_tensor = torch.from_numpy(features_np_array).type(torch.FloatTensor)
root = Tree(features_tensor)
root.action = Action(
action_id=0,
action_name="current_state",
action_value=0,
cost_value=0,
feature_number=0
) #We create a fictive action for the root, just to make sure the algorithm runs well. We will delete this from the proposed list.
net = main.Net(DROPOUT_RATE=0.1)
net.load_state_dict(torch.load('models/final_weights.pkl', map_location='cpu'))
res = monte_carlo_tree_search(root)
import pandas as pd
import torch
import main
# Hyper parameters
BATCH_SIZE = 20
LEARNING_RATE = 0.001 # The optimal learning rate for the Adam optimizer
BEST_EPOCH = 5
DROPOUT_RATE = 0.1
val_dataset = main.Dataset('dataset/prep_unbiased/val.txt')
val_dataloader = torch.utils.data.DataLoader(val_dataset, batch_size=BATCH_SIZE)
net = main.Net(DROPOUT_RATE)
net.load_state_dict(torch.load('models/split_10_90_batchsize_' + str(BATCH_SIZE) + '_lr_' + str(LEARNING_RATE) + '_dropout_'+ str(DROPOUT_RATE) + '_epoch_' + str(BEST_EPOCH) +'.pkl'))
print(main.compute_loss(net, val_dataloader))
print()
def hello():
print('hello called')
origination_channel = request.args.get('origination_channel', type=float)
seller_name = request.args.get('seller_name', type=float)
interest_rate = request.args.get('interest_rate', type=float)
upb = request.args.get('upb', type=float)
orig_loan_t = request.args.get('orig_loan_t', type=float)
total_price = request.args.get('total_price', type=float)
first_lien = request.args.get('first_lien', type=float)
current_amount = request.args.get('current_amount', type=float)
second_amount = request.args.get('second_amount', type=float)
num_borr = request.args.get('num_borr', type=float)
monthly_payments = request.args.get('monthly_payments', type=float)
income = request.args.get('income', type=float)
borrower_credit_score = request.args.get('borrower_credit_score',
type=float)
first_time = request.args.get('first_time', type=float)
loan_purp = request.args.get('loan_purp', type=float)
num_units = request.args.get('num_units', type=float)
occ_type = request.args.get('occ_type', type=float)
zip = request.args.get('zip', type=float)
co_credit_score = request.args.get('co_credit_score', type=float)
ins_perc = request.args.get('ins_perc', type=float)
ins_type = request.args.get('ins_type', type=float)
reloc_ind = request.args.get('reloc_ind', type=float)
state = request.args.get('state', type=float)
stats = pd.read_csv('dataset\statistics.csv')
stats_mean = stats.iloc[1]
stats_std = stats.iloc[2]
# Other parameters we need to calculate:
ltv = upb / total_price * 100
cltv = (first_lien + current_amount + second_amount) / total_price * 100
dti = monthly_payments / income * 100
if interest_rate == -1:
interest_rate = stats_mean[3]
# Build feature list for new person
new_person = np.array([
origination_channel, seller_name, interest_rate, upb, orig_loan_t, ltv,
cltv, num_borr, dti, borrower_credit_score, first_time, loan_purp,
num_units, occ_type, 1.0, state, zip, ins_perc, co_credit_score,
ins_type, reloc_ind
])
# Instansiate net
net = main.Net(DROPOUT_RATE=0.0001)
net.load_state_dict(
torch.load('models/final_weights.pkl', map_location='cpu'))
# Call the net with the person's online data
new_person_np = new_person
new_person = pd.DataFrame(new_person).transpose().astype(float)
new_person_normalized = preprocessor.norm_features(new_person,
montecarlo.stats)
new_person_normalized = torch.tensor(new_person_normalized.values).float()
net_out = float(net.forward(new_person_normalized))
# Either tell the client his application will be approved, or call monte-carlo code to suggest
print(net_out)
if round(net_out) == 0:
print("Mortgage request is approved")
return str("Mortgage request is approved")
else:
print("Mortgage request is declined")
features_tensor = torch.from_numpy(new_person_np).type(
torch.FloatTensor)
root = montecarlo.Tree(features_tensor)
root.action = montecarlo.Action(
action_id=0,
action_name="current_state",
action_value=0,
cost_value=0,
feature_number=0
) # We create a fictive action for the root, just to make sure the algorithm runs well. We will delete this from the proposed list.
plan = montecarlo.monte_carlo_tree_search(root)
return plan
# -*- coding: utf-8 -*-#
# ---------------------------------------------
# Name: testModule
# Description:
# Author: Laity
# Date: 2021/9/29
# ---------------------------------------------
import torch
import main
import getData
import matplotlib.pyplot as plt
net = main.Net()
dic = torch.load('net.pt')
net.load_state_dict(dic)
testData, testLabel = getData.getTestData()
cha = [
'零', '一', '二', '三', '四', '五', '六', '七', '八', '九', '十', '百', '千', '万', '亿'
]
acc = 0
for i in range(getData.test_size): # len(testLabel)
outputs = net(testData[i])
res = 0
for j in range(15):
if outputs[0][j] == max(outputs[0]):
res = j
break
print('=======================')