from models import ModelBuilder, activate from utils import AverageMeter, \ recover_rgb, magnitude2heatmap, \ istft_reconstruction, warpgrid, \ combine_video_audio, save_video, makedirs from viz import plot_loss_metrics, HTMLVisualizer import imageio def main(args): dataset_train = MUSICMixDataset(args.list_train, args, split='train') loader_train = torch.utils.data.DataLoader(dataset_train, batch_size=20, shuffle=True, num_workers=0, drop_last=True) print("a") for data in loader_train: print('new iter') print(data) print("b") if __name__ == '__main__': parser = ArgParser() args = parser.parse_train_arguments() args.batch_size = args.num_gpus * args.batch_size_per_gpu args.device = torch.device("cuda") main(args)
from torch import nn from arguments import ArgParser from constants import * from data import * from logger import * import models from training_tools import Evaluator, Optimizer, Predictor from training_tools import find_save_dir, save_progress, load_progress from utils import * # Define cloud storage here cloudFS = GCStorage.get_CloudFS(PROJECT_ID, GC_BUCKET, CREDENTIAL_PATH) # Parse command line arguments parser = ArgParser() args = parser.parse_args() # Verify arguments # check_args(args) # Setup logger logger = args.misc_args.logger # Setup device device = args.misc_args.device # Select columns selected_columns = sorted(select_columns(args.data_args.data_spec)) if args.model_args.wavelet: args.model_args.input_channels = len(selected_columns) * 2