args module

class args.BackwardStepConfig(parallel: bool = True, n_gpu: int = 4, ntrain: int = 48, ntest: int = 4, noise_std: float = 0.05, training_data_dir: str = '../step-training/', testing_data_dir: str = '../step-testing/', beta: float = 200, dx: float = 0.03125, dy: float = 0.03125, nu: float = 0.005, batch_size: int = 32, test_batch_size: int = 4, max_grad_norm: float = 0.01, nic: int = 4, glow_upscale: int = 2)

Bases: object

Configuration Dataclass to setup the model for the backward-step numerical test case.

batch_size: int = 32

beta: float = 200

dx: float = 0.03125

dy: float = 0.03125

glow_upscale: int = 2

max_grad_norm: float = 0.01

n_gpu: int = 4

nic: int = 4

noise_std: float = 0.05

ntest: int = 4

ntrain: int = 48

nu: float = 0.005

parallel: bool = True

test_batch_size: int = 4

testing_data_dir: str = '../step-testing/'

training_data_dir: str = '../step-training/'

class args.CylinderConfig(parallel: bool = True, n_gpu: int = 4, ntrain: int = 96, ntest: int = 4, noise_std: float = 0.05, training_data_dir: str = '../cylinder-training/', testing_data_dir: str = '../cylinder-testing/', beta: float = 200, dx: float = 0.078125, dy: float = 0.078125, nu: float = 0.005, batch_size: int = 64, test_batch_size: int = 4, max_grad_norm: float = 1.0, nic: int = 3, glow_upscale: int = 4)

Bases: object

Configuration Dataclass to setup the model for the cylinder-array numerical test case.

batch_size: int = 64

beta: float = 200

dx: float = 0.078125

dy: float = 0.078125

glow_upscale: int = 4

max_grad_norm: float = 1.0

n_gpu: int = 4

nic: int = 3

noise_std: float = 0.05

ntest: int = 4

ntrain: int = 96

nu: float = 0.005

parallel: bool = True

test_batch_size: int = 4

testing_data_dir: str = '../cylinder-testing/'

training_data_dir: str = '../cylinder-training/'

class args.Parser

Bases: argparse.ArgumentParser

Program arguments, only a few are listed in the documentation.

Parameters

• exp-dir (string) – Directory to save experiments

• exp-type (string) – Experiment type

• parallel (bool) – Use parallel GPUs for training

• n_gpu (int) – Number of GPUs to use for training, defaults to 1

• training_data_dir (string) – File directory to training data

• testing_data_dir (string) – File directory to testing data

• ntrain (int) – Number of training data

• ntest (int) – Number of testing data

• epoch_start (int) – Epoch to start at, will load pre-trained network

• epochs (int) – Number of epochs to train

• lr (float) – ADAM optimizer learning rate

Note

Use python main.py –help for more information. Only several key of arguments are listed here.

loadConfig(args)

Loads experimental configurations.

mkdirs(*directories)

Makes a directory if it does not exist

parse(dirs=True)

Parse program arguments Args: dirs (boolean): True to make file directories for predictions and models