CovaMNet_Train_5way1shot py 程序分块儿解读

CovaMNet_Train_5way1shot.py 程序分块儿解读

import 部分

基本import

import argparse
import os
import random
import shutil
import numpy as np
import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.optim as optim
import torch.utils.data
import torchvision.datasets as dset
import torchvision.transforms as transforms
import torchvision.utils as vutils
from torch.autograd import grad
import time
from torch import autograd
from PIL import ImageFile
import scipy as sp
import scipy.stats
import sys
sys.dont_write_bytecode = True

自己编写的两部分

• dataset文件下的datasets_csv.py
• models文件下的network.py

(1) from dataset.datasets_csv import Imagefolder_csv

(2)import models.network as CovaNet

Linux参数设置

ImageFile.LOAD_TRUNCATED_IMAGES = True
os.environ['CUDA_DEVICE_ORDER']='PCI_BUS_ID'
os.environ['CUDA_VISIBLE_DEVICES']='0'

parser = argparse.ArgumentParser()
parser.add_argument('--dataset_dir', default='/Datasets/miniImageNet--ravi', help='the path of the data')
parser.add_argument('--data_name', default='miniImageNet', help='miniImageNet|StanfordDog|StanfordCar|CubBird')
parser.add_argument('--mode', default='test', help='train|val|test')
parser.add_argument('--outf', default='./results/CovaMNet')
parser.add_argument('--resume', default='', type=str, help='path to the lastest checkpoint (default: none)')
parser.add_argument('--basemodel', default='Conv64', help='Conv64')
parser.add_argument('--workers', type=int, default=8)
#  Few-shot parameters  #
parser.add_argument('--imageSize', type=int, default=84)
parser.add_argument('--episodeSize', type=int, default=1, help='the mini-batch size of training')
parser.add_argument('--testepisodeSize', type=int, default=1, help='one episode is taken as a mini-batch')
parser.add_argument('--epochs', type=int, default=30, help='the total number of training epoch')
parser.add_argument('--episode_train_num', type=int, default=10000, help='the total number of training episodes')
parser.add_argument('--episode_val_num', type=int, default=1000, help='the total number of evaluation episodes')
parser.add_argument('--episode_test_num', type=int, default=600, help='the total number of testing episodes')
parser.add_argument('--way_num', type=int, default=5, help='the number of way/class')
parser.add_argument(<