Olga (Ge Ya) Xu
- torchvision==0.2.0
- matplotlib==2.2.2
- six==1.11.0
- torch.egg==info
- tqdm==4.23.0
- numpy==1.14.2
- Pillow==5.3.0
- torch==0.4.1.post2
- Project writeup: overleaf link
git clone https://github.com/oooolga/IFT6390_Project.git
cd IFT6390_Project
mkdir saved_models results
python main.py ...
usage: main.py [-h] [-lr LEARNING_RATE] [--batch_size BATCH_SIZE]
[--test_batch_size TEST_BATCH_SIZE] [--epochs EPOCHS]
[--seed SEED] [--weight_decay WEIGHT_DECAY] --model_name
MODEL_NAME [--load_model LOAD_MODEL] [--optimizer {Adam,SGD}]
[--dataset {CIFAR,FMNIST,EMNIST}] [--model {CNN,NN,Regression}]
[--plot_freq PLOT_FREQ]
optional arguments:
-h, --help show this help message and exit
-lr LEARNING_RATE, --learning_rate LEARNING_RATE
Learning rate.
--batch_size BATCH_SIZE
Mini-batch size for training.
--test_batch_size TEST_BATCH_SIZE
Mini-batch size for testing.
--epochs EPOCHS Total number of epochs.
--seed SEED Random number seed.
--weight_decay WEIGHT_DECAY
Weight decay.
--model_name MODEL_NAME
Model name.
--load_model LOAD_MODEL
Load model path.
--optimizer {Adam,SGD}
Optimizer type.
--dataset {CIFAR,FMNIST,EMNIST}
Dataset choice.
--model {CNN,NN,Regression}
Model type.
--plot_freq PLOT_FREQ
plot_freq
python main.py -lr 0.01 --epochs 100 --weight_decay 5e-4 --model_name CIFAR_CNN --model CNN --optimizer SGD --dataset CIFAR
Model type: CNN
Dataset: CIFAR
Optimizer type: SGD
Learning rate: 0.01
Total number of epochs: 100
Learning rate: 0.01
Weight decay: 0.0005
Batch size: 50
Plot frequency: 5
python main.py -lr 0.01 --epochs 100 --weight_decay 5e-4 --model_name CIFAR_CNN --model CNN --optimizer SGD --dataset CIFAR --load_model saved_models/CIFAR_CNN.pt
python evaluate.py ...
usage: evaluate.py [-h] [--dataset {CIFAR,FMNIST,EMNIST}]
[--model {CNN,NN,Regression}] --load_model LOAD_MODEL
[--batch_size BATCH_SIZE]
optional arguments:
-h, --help show this help message and exit
--dataset {CIFAR,FMNIST,EMNIST}
Dataset choice.
--model {CNN,NN,Regression}
Model type.
--load_model LOAD_MODEL
Load model path.
--batch_size BATCH_SIZE
Mini-batch size for testing.
python evaluate.py --dataset CIFAR --model CNN --load_model saved_models/CIFAR_CNN.pt