main_TO.py

import argparse
import torch
import random
import time
from datetime import datetime
from onescience.models.GPs import GPPLUS
from onescience.utils.GP_TO import set_seed,get_data_fluid
from train import find_TO

def main(args):
    tkwargs = {
        "dtype": torch.float,
        "device": torch.device(f"cuda:{args.gpu}" if torch.cuda.is_available() else "cpu"),
    }

    # Fix Seed
    random_states = 11

    problem = args.problem  # 从参数获取problem

    ############################### Generate Data ##############################################
    set_seed(random_states)

    NOW = datetime.now()
    d = NOW.strftime("%B%d_%H-%M")

    N_train_per_BC = 25
    N_col_domain = 10000
    X_col, X_train, Sol_train = get_data_fluid(problem=problem, N_col_domain=N_col_domain,
                                              N_train=N_train_per_BC)
    mean_function = 'neural_network'  
    NN_archi =[64,64,64,64,64]

    # Names of the outputs 
    output_names = ['u', 'v', 'p', 'ro']

    # Define all models as a list of models
    models = [
        GPPLUS(
            train_x=X_train[i].type(tkwargs["dtype"]),
            train_y=Sol_train[i].type(tkwargs["dtype"]),
            collocation_x=X_col.type(tkwargs["dtype"]).clone().requires_grad_(True).to(tkwargs["device"]),
            basis=mean_function,
            NN_layers_base=NN_archi,
            name_output=name,
            device=tkwargs["device"],
            dtype=tkwargs["dtype"]
        ).to(**tkwargs)
        for i, name in enumerate(output_names)
    ]

    ############################### Fit Model ##########################################
    start_time = time.time()  # get the current time
    loss_history = find_TO(model_list=models,
                           num_iter=50000,
                           lr_default=0.001,
                           title=f"seed{random_states}_{problem}_{d}",
                           problem=problem,
                           diff_method='Numerical')
    end_time = time.time()

    print(f'The total time in seconds is {end_time - start_time}')

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Run TO model training")
    parser.add_argument('--problem', type=str, default='doublepipe',
                        choices=['doublepipe', 'diffuser', 'rugby', 'pipebend'],
                        help="Problem name to select the dataset and configuration")
    parser.add_argument('--gpu', type=int, default=0,
                        help="GPU id to use, default=0")
    args = parser.parse_args()

    main(args)