-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_backbone.py
39 lines (27 loc) · 1.36 KB
/
train_backbone.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
import argparse
from src.backbone import BackboneNetwork, get_MNIST
from src.training import TrainerBackbone
import torch
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument("--name", type=str, default="backbone_model.pth",
help="Name of the model to save. Default backbone_model.pth")
parser.add_argument("--epochs", type=int, default=20,
help="Number of epochs to train the model. Default: 20")
parser.add_argument("--batch_size", type=int, default=64,
help="Batch size. Default: 64")
parser.add_argument("--keep_best", action="store_true",
help="Keep the best model during training")
return parser.parse_args()
if __name__ == '__main__':
args = parse_args()
device = "cuda" if torch.cuda.is_available() else "cpu"
train_dataloader, test_dataloader = get_MNIST(batch_size=args.batch_size)
model = BackboneNetwork().to(device)
loss = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters())
trainer = TrainerBackbone(model, optimizer, loss, device)
print("Training Backbone model...")
vals = trainer.train(train_dataloader, test_dataloader, epochs=args.epochs,
keep_best=args.keep_best)
trainer.save(f"models/{args.name}")