Levy_Ito_Model

Environment Setting

First, navigate into the directory.

cd Levy_Ito_Model

Next, install the necessary packages.

pip install -r requirements.txt

Training Command

To train the model, use the following command:

CUDA_VISIBLE_DEVICES=0,1 LOCAK_RANK=0,1 torchrun --master_port 12355 --nproc_per_node 2 main.py --exp t_cifar10_2.0  --seed 0 --config cifar10_ddpm.yaml --ddp 

Parameters:

• --exp [folder_name]: Name of the folder to save the experiment results.
• --seed [integer]: Seed for reproducibility.
• --config [configuration_file]: Configuration file to reference. Example files (cifar10_ddpm.yaml and cifar10_ncnsnpp.yaml) are located in the config directory.
• --ddp: Run the code in a GPU environment.
• --resume: Resume training from a saved checkpoint in the specified folder.

Sampling Command

To sample from the trained model, use the following command:

CUDA_VISIBLE_DEVICES=0,1 LOCAK_RANK=0,1 torchrun --master_port 12355 --nproc_per_node 2 main.py --exp t_cifar10_2.0  --seed 0 --config cifar10_ddpm.yaml --ddp --resume --sampling 

Parameters:

• --exp [folder_name]: Name of the folder containing the trained model.
• --seed [integer]: Seed for reproducibility.
• --config [configuration_file]: Configuration file to reference. Example files (cifar10_ddpm.yaml and cifar10_ncnsnpp.yaml) are located in the config directory.
• --ddp: Run the code in a GPU environment.
• --resume: Load the saved checkpoint from the specified folder for sampling.
• --sample: Enable sampling mode.
• --fid: Calculate the Frechet Inception Distance (FID).