forked from kialanpillay/cogspms
-
Notifications
You must be signed in to change notification settings - Fork 0
/
user_model_managers.py
307 lines (263 loc) · 13.6 KB
/
user_model_managers.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
from stdnn.experiments.results import RunResult
from stdnn.models.manager import STModelManager
class GWNManager(STModelManager):
def __init__(self):
super().__init__()
def run_pipeline(self, config):
"""
Executes the machine learning pipeline for the given model
Parameters
----------
config : ExperimentConfig
An ExperimentConfig object containing the parameters for the model and pipeline
Returns
-------
results : RunResult
A RunResult containing the results collected in the pipeline
"""
train, valid, test, train_scaler, test_scaler = self.preprocess(**config.get_preprocessing_params())
train_results = self.train_model(train, valid, train_scaler, **config.get_training_params())
test_results = self.test_model(test, test_scaler, **config.get_testing_params())
result = RunResult(
{**train_results, **test_results}
)
return result
from user_preprocess import preprocess
from user_train import train_model
from user_validate import validate_model
from user_test import test_model
from user_predict import predict as _custom_inference
# class LSTM(nn.Module):
# """
# A baseline 100-layer LSTM model with a single output layer for univariate predictions
# """
# def __init__(self, input_size=1, hidden_layers=100, output_size=1):
# """
# Parameters
# ----------
# input_size : int, optional
# Input layer dimension
# hidden_layers : int, optional
# Number of hidden layers
# output_size : int, optional
# Output layer dimension
# """
# super().__init__()
# self.hidden_layers = hidden_layers
# self.lstm = nn.LSTM(input_size=input_size, hidden_size=hidden_layers)
# self.linear = nn.Linear(hidden_layers, output_size)
# self.hidden_cell = (torch.zeros(1, 1, self.hidden_layers),
# torch.zeros(1, 1, self.hidden_layers))
# def forward(self, in_seq):
# lstm_out, self.hidden_cell = self.lstm(in_seq.view(len(in_seq), 1, -1), self.hidden_cell)
# forecast = self.linear(lstm_out.view(len(in_seq), -1))
# return forecast
# class LSTMManager(STModelManager):
# def __init__(self):
# super().__init__()
# @timed(operation_name="Train")
# def train_model(self, train_data, valid_data, args, result_file):
# """
# Trains a LSTM model and returns a set of validation performance metrics
# Parameters
# ----------
# train_data : numpy.ndarray
# Train set
# valid_data : numpy.ndarray
# Validation set
# args : argparse.Namespace
# Command line arguments
# result_file : str
# Directory to store trained model parameter files
# Returns
# -------
# dict
# """
# self.model.to(args.device)
# if len(train_data) == 0:
# raise Exception('Cannot organize enough training data')
# if len(valid_data) == 0:
# raise Exception('Cannot organize enough validation data')
# if args.norm_method == 'z_score':
# train_mean = np.mean(train_data[:, args.lstm_node], axis=0)
# train_std = np.std(train_data[:, args.lstm_node], axis=0)
# norm_statistic = {"mean": [train_mean], "std": [train_std]}
# elif args.norm_method == 'min_max':
# train_min = np.min(train_data[:, args.lstm_node], axis=0)
# train_max = np.max(train_data[:, args.lstm_node], axis=0)
# norm_statistic = {"min": [train_min], "max": [train_max]}
# else:
# norm_statistic = None
# if norm_statistic is not None:
# with open(os.path.join(result_file, 'norm_stat.json'), 'w') as f:
# json.dump(norm_statistic, f)
# if args.optimizer == 'RMSProp':
# optimizer = torch.optim.RMSprop(params=self.model.parameters(), lr=args.lr)
# elif args.optimizer == 'SGD':
# optimizer = torch.optim.SGD(params=self.model.parameters(), lr=args.lr)
# elif args.optimizer == 'Adagrad':
# optimizer = torch.optim.Adagrad(params=self.model.parameters(), lr=args.lr)
# elif args.optimizer == 'Adadelta':
# optimizer = torch.optim.Adadelta(params=self.model.parameters(), lr=args.lr)
# else:
# optimizer = torch.optim.Adam(params=self.model.parameters(), lr=args.lr, betas=(0.9, 0.999))
# lr_scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer=optimizer, gamma=args.decay_rate)
# train_set = user_preprocessing.loader.ForecastDataset(train_data, window_size=args.window_size,
# horizon=args.horizon, normalize_method=args.norm_method,
# norm_statistic=norm_statistic)
# valid_set = user_preprocessing.loader.ForecastDataset(valid_data, window_size=args.window_size,
# horizon=args.horizon, normalize_method=args.norm_method,
# norm_statistic=norm_statistic)
# train_loader = torch.utils.data.DataLoader(train_set, batch_size=args.batch_size, drop_last=False, shuffle=True,
# num_workers=0)
# valid_loader = torch.utils.data.DataLoader(valid_set, batch_size=args.batch_size, shuffle=False, num_workers=0)
# criterion = nn.MSELoss(reduction='mean').to(args.device)
# total_params = 0
# for name, parameter in self.model.named_parameters():
# if not parameter.requires_grad:
# continue
# param = parameter.numel()
# total_params += param
# print(f"Total Trainable Params: {total_params}")
# print("LSTM")
# print()
# best_validate_mae = np.inf
# validate_score_non_decrease_count = 0
# performance_metrics = {}
# for epoch in range(args.epoch):
# epoch_start_time = time.time()
# self.model.train()
# loss_total = 0
# cnt = 0
# for i, (inputs, target) in enumerate(train_loader):
# inputs = inputs.to(args.device)
# target = target.to(args.device)
# optimizer.zero_grad()
# self.model.hidden_cell = (torch.zeros(1, 1, self.model.hidden_layers).to(args.device),
# torch.zeros(1, 1, self.model.hidden_layers).to(args.device))
# forecast = self.model(inputs[:, :, args.lstm_node])
# loss = criterion(forecast, target[:, :, args.lstm_node])
# loss.backward()
# cnt += 1
# optimizer.step()
# loss_total += float(loss)
# print('Epoch {:2d} | Time: {:4.2f}s | Total Loss: {:5.4f}'.format(epoch + 1, (
# time.time() - epoch_start_time), loss_total / cnt))
# self.save_model(result_file, epoch)
# if (epoch + 1) % args.exponential_decay_step == 0:
# lr_scheduler.step()
# if (epoch + 1) % args.validate_freq == 0:
# is_best = False
# print('------ VALIDATE ------')
# performance_metrics = \
# self.validate_model(args.lstm_node, valid_loader, args.device, args.norm_method, norm_statistic)
# if args.horizon == 1:
# self.validate_model(args.lstm_node, valid_loader, args.device, args.norm_method, norm_statistic,
# True)
# if np.abs(best_validate_mae) > np.abs(performance_metrics['mae']):
# best_validate_mae = performance_metrics['mae']
# is_best = True
# validate_score_non_decrease_count = 0
# else:
# validate_score_non_decrease_count += 1
# if is_best:
# self.save_model(result_file)
# if args.early_stop and validate_score_non_decrease_count >= args.early_stop_step:
# break
# return performance_metrics
# def validate_model(self, node, data_loader, device, norm_method, statistic, naive=False):
# """
# Validates a LSTM or naive model and returns raw and normalized error metrics
# computed on validation set predictions
# Parameters
# ----------
# model : Union[GraphWaveNet, MTGNN, Model]
# Graph neural network model for validation
# node: int
# Index of node to forecast
# data_loader : torch.Dataset
# An iterable dataset
# device : str
# Torch device
# norm_method: str
# Raw data normalization method
# statistic: dict
# Raw data statistics
# naive: bool
# Compute last-value (naive) model performance measures
# Returns
# -------
# dict
# """
# forecast_set = []
# target_set = []
# if naive:
# for i, (inputs, target) in enumerate(data_loader):
# forecast_set.append(inputs[:, -1, node])
# target_set.append(target[:, :, node])
# else:
# self.model.eval()
# with torch.no_grad():
# for i, (inputs, target) in enumerate(data_loader):
# inputs = torch.Tensor(inputs[:, :, node]).to(device)
# target_norm = torch.Tensor(target[:, :, node]).to(device)
# self.model.hidden = (torch.zeros(1, 1, self.model.hidden_layers),
# torch.zeros(1, 1, self.model.hidden_layers))
# forecast_result = self.model(inputs)
# forecast_set.append(forecast_result.squeeze())
# target_set.append(target_norm.detach().cpu().numpy())
# forecast_norm = torch.cat(forecast_set, dim=0)[:np.concatenate(target_set, axis=0).shape[0], ...].detach().cpu() \
# .numpy()
# target_norm = np.concatenate(target_set, axis=0)
# if target_norm.shape[1] == 1:
# target_norm = target_norm[:, 0]
# if norm_method == 'min_max':
# scale = statistic['max'] - statistic['min'] + 1e-8
# forecast = forecast_norm * scale + statistic['min']
# target = target_norm * scale + statistic['min']
# elif norm_method == 'z_score':
# forecast = forecast_norm * statistic['std'] + statistic['mean']
# target = target_norm * statistic['std'] + statistic['mean']
# else:
# forecast, target = forecast_norm, target_norm
# score = evaluate(target, forecast)
# score_norm = evaluate(target_norm, forecast_norm)
# if naive:
# print("LAST VALUE MODEL")
# print("NORM - MAPE {:>8.4f}% | MAE {:>10.4f} | RMSE {:>10.4f}".format(score_norm[0] * 100, score_norm[1],
# score_norm[2]))
# print("RAW - MAPE {:>8.4f}% | MAE {:>10.4f} | RMSE {:>10.4f}".format(score[0] * 100, score[1], score[2]))
# return dict(mae=score[1], mape=score[0], rmse=score[2])
# @timed(operation_name="Evaluation")
# def test_model(self, test_data, args, result_train_file):
# """
# Evaluates a LSTM model and returns raw and normalized error metrics
# computed on out-of-sample set predictions
# Parameters
# ----------
# test_data : numpy.ndarray
# Test set
# args : argparse.Namespace
# Command line arguments
# result_train_file : str
# Directory to load trained model parameter files
# """
# with open(os.path.join(result_train_file, 'norm_stat.json'), 'r') as f:
# normalize_statistic = json.load(f)
# if not self.has_model():
# self.load_model(result_train_file)
# test_set = user_preprocessing.loader.ForecastDataset(test_data, window_size=args.window_size, horizon=args.horizon,
# normalize_method=args.norm_method)
# test_loader = torch.utils.data.DataLoader(test_set, batch_size=args.batch_size, drop_last=False,
# shuffle=False, num_workers=0)
# performance_metrics = self.validate_model(args.lstm_node, test_loader, args.device, args.norm_method,
# normalize_statistic)
# mae, mape, rmse = performance_metrics['mae'], performance_metrics['mape'], performance_metrics['rmse']
# print('Test Set Performance: MAPE: {:5.2f} | MAE: {:5.2f} | RMSE: {:5.2f}'.format(mape * 100, mae, rmse))
# if args.horizon == 1:
# performance_metrics = self.validate_model(args.lstm_node, test_loader, args.device, args.norm_method,
# normalize_statistic, True)
# mae, mape, rmse = performance_metrics['mae'], performance_metrics['mape'], performance_metrics['rmse']
# print(
# 'Last-Value Test Set Performance: MAPE: {:5.2f} | MAE: {:5.2f} | RMSE: {:5.2f}'.format(mape * 100, mae,
# rmse))