-
Notifications
You must be signed in to change notification settings - Fork 0
/
optuna_optimizer.py
70 lines (58 loc) · 2.25 KB
/
optuna_optimizer.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
import json
import optuna
import torch
import pytorch_lightning as pl
from pytorch_forecasting import TemporalFusionTransformer, QuantileLoss
from pytorch_lightning.callbacks import EarlyStopping
# Load existing configuration
with open('config.json', 'r') as config_file:
config = json.load(config_file)
# Extract trainer parameters
trainer_params = config["trainer"]
def objective(trial, training, train_dataloader, val_dataloader):
# Define hyperparameter search space
learning_rate = trial.suggest_float("learning_rate", 1e-5, 1e-2, log=True)
hidden_size = trial.suggest_categorical("hidden_size", [8, 16, 32])
dropout = trial.suggest_float("dropout", 0.1, 0.5)
attention_head_size = trial.suggest_categorical("attention_head_size", [1, 2, 4])
hidden_continuous_size = trial.suggest_categorical("hidden_continuous_size", [8, 16, 32])
# Model
model = TemporalFusionTransformer.from_dataset(
training,
learning_rate=learning_rate,
hidden_size=hidden_size,
attention_head_size=attention_head_size,
dropout=dropout,
hidden_continuous_size=hidden_continuous_size,
loss=QuantileLoss(),
log_interval=10
)
# Trainer setup with loaded parameters
trainer = pl.Trainer(
**trainer_params,
accelerator='gpu' if torch.cuda.is_available() else 'cpu',
callbacks=[EarlyStopping(monitor="val_loss", patience=3, mode="min")]
)
# Train the model
trainer.fit(
model,
train_dataloaders=train_dataloader,
val_dataloaders=val_dataloader
)
# Objective: the best validation loss
return trainer.callback_metrics["val_loss"].item()
study = optuna.create_study(direction="minimize")
study.optimize(objective, n_trials=20) # Adjust the number of trials as needed
# Extract the best parameters for TFT
best_tft_params = study.best_trial.params
# Update the config dictionary
config["tft"] = best_tft_params
# Save updated configuration to JSON file
with open('train_config_optuna.json', 'w') as config_file:
json.dump(config, config_file, indent=4)
# Print the best trial results
print("Best trial:")
print(f"Value: {study.best_trial.value}")
print("Params: ")
for key, value in best_tft_params.items():
print(f" {key}: {value}")