135 lines
4.7 KiB
Python
135 lines
4.7 KiB
Python
# -*- encoding:utf-8 -*-
|
|
|
|
'''
|
|
@Author : dingjiawen
|
|
@Date : 2023/11/10 16:27
|
|
@Usage :
|
|
@Desc :
|
|
'''
|
|
|
|
import numpy as np
|
|
import torch
|
|
from RUL.otherIdea.dctChannelEmbedLSTM.loadData import getDataset, getTotalData
|
|
from RUL.otherIdea.dctChannelEmbedLSTM.model import PredictModel
|
|
from torch.utils.data import DataLoader
|
|
import matplotlib.pyplot as plt
|
|
|
|
|
|
def plot_prediction(total_data, predicted_data_easy, predicted_data_hard):
|
|
pic1 = plt.figure(figsize=(8, 6), dpi=200)
|
|
|
|
'''保存的模型参数的路径'''
|
|
|
|
from matplotlib import rcParams
|
|
|
|
config = {
|
|
"font.family": 'Times New Roman', # 设置字体类型
|
|
"axes.unicode_minus": False, # 解决负号无法显示的问题
|
|
"axes.labelsize": 13
|
|
}
|
|
rcParams.update(config)
|
|
|
|
# 简单预测图
|
|
plt.subplot(2, 1, 1)
|
|
plt.plot(total_data)
|
|
plt.plot(predicted_data_easy)
|
|
plt.title('Easy Prediction')
|
|
plt.xlabel('time')
|
|
plt.ylabel('loss')
|
|
# plt.legend(loc='upper right')
|
|
|
|
# 困难预测图
|
|
plt.subplot(2, 1, 2)
|
|
plt.plot(total_data)
|
|
plt.plot(predicted_data_hard)
|
|
plt.title('Easy Prediction')
|
|
plt.xlabel('time')
|
|
plt.ylabel('loss')
|
|
|
|
# plt.legend(loc='upper right')
|
|
|
|
|
|
# plt.scatter()
|
|
|
|
plt.show()
|
|
|
|
|
|
# 仅使用预测出来的最新的一个点预测以后
|
|
def predictOneByOne(model, train_data, predict_num=50):
|
|
# 取出训练数据的最后一条
|
|
each_predict_data = train_data
|
|
predicted_list = np.empty(shape=(predict_num, 1)) # (5,filter_num,30)
|
|
# all_data = total_data # (1201,)
|
|
for each_predict in range(predict_num):
|
|
# predicted_data.shape : (1,1)
|
|
predicted_data = model(each_predict_data).cpu().detach().numpy()[-1] # (batch_size,filer_num,1)
|
|
predicted_list[each_predict] = predicted_data
|
|
each_predict_data = each_predict_data.numpy()
|
|
# (1,1) => (10,1)
|
|
# 中间拼接过程： (1) => (10) => (40,10) => (30,40,10)
|
|
a = np.concatenate([each_predict_data[-1, -1, 1:], predicted_data], axis=0)
|
|
b = np.concatenate([each_predict_data[-1, 1:, :], np.expand_dims(a, axis=0)], axis=0)
|
|
c = np.concatenate([each_predict_data[1:, :, :], np.expand_dims(b, axis=0)], axis=0)
|
|
|
|
each_predict_data = torch.tensor(c)
|
|
|
|
return np.squeeze(predicted_list)
|
|
|
|
|
|
def test(hidden_num, feature, predict_num, batch_size, save_path, is_single=True, is_norm=False):
|
|
total_data, total_dataset = getTotalData(hidden_num, feature, is_single=is_single, is_norm=is_norm)
|
|
train_dataset, val_dataset = getDataset(hidden_num, feature, predict_num=predict_num, is_single=is_single,
|
|
is_norm=is_norm)
|
|
|
|
train_loader = DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=False)
|
|
val_loader = DataLoader(dataset=val_dataset, batch_size=batch_size, shuffle=False)
|
|
|
|
# 加载网络
|
|
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
|
model = PredictModel(input_dim=feature).to(device)
|
|
|
|
model.load_state_dict(
|
|
torch.load(save_path, map_location=device)
|
|
)
|
|
|
|
print(model)
|
|
params_num = sum(param.numel() for param in model.parameters())
|
|
print('参数数量：{}'.format(params_num))
|
|
|
|
model.eval()
|
|
|
|
predicted_data_easy = total_data[:hidden_num + feature, ]
|
|
predicted_data_hard = total_data[:hidden_num + feature, ]
|
|
|
|
with torch.no_grad():
|
|
for batch_idx, (data, label) in enumerate(train_loader):
|
|
data, label = data.to(device), label.to(device)
|
|
last_train_data = data
|
|
each_predicted_data = torch.squeeze(model(data)).cpu().detach().numpy()
|
|
predicted_data_easy = np.concatenate(
|
|
[predicted_data_easy, each_predicted_data],
|
|
axis=0)
|
|
predicted_data_hard = np.concatenate(
|
|
[predicted_data_hard, each_predicted_data],
|
|
axis=0)
|
|
|
|
# 简单版的，每次预测重新用已知的
|
|
for batch_idx, (data, label) in enumerate(val_loader):
|
|
data, label = data.to(device), label.to(device)
|
|
each_predicted_data = torch.squeeze(model(data)).cpu().detach().numpy()
|
|
predicted_data_easy = np.concatenate(
|
|
[predicted_data_easy, each_predicted_data],
|
|
axis=0)
|
|
|
|
# 困难版的,每次预测基于上次的预测
|
|
predicted_data_hard = np.concatenate([predicted_data_hard,
|
|
predictOneByOne(model, last_train_data, predict_num=predict_num)], axis=0)
|
|
|
|
plot_prediction(total_data, predicted_data_easy, predicted_data_hard)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
test(40, 10, 50, 32,
|
|
"./parameters\dctLSTM_hidden40_feature10_predict50_epoch80_trainLoss0.05818896507844329_valLoss0.21667905896902084.pkl"
|
|
)
|