# Global environment setup.
import os
# Arg parser
from config import globalparser
args = vars(globalparser.getparser().parse_args())
from config import globalconfig
globalconfig.run(args, False)
import glob
import numpy as np
def print_prompt():
available_data = np.array([])
datalist = glob.glob(os.environ['datapath'] + '/*')
for p in datalist:
if os.path.exists(p + '/info.json'):
available_data = np.append(available_data, p)
available_data = [x[x.rfind('/') + 1:] for x in available_data]
print('Available datas:')
print(available_data)
modellist = glob.glob('./Loader/Model/*.py')
modellist = [x[x.rfind('/') + 1:x.rfind('.py')] for x in modellist]
print('Available models:')
print(modellist)
print_prompt()
# Global environment setup.
import os
# Arg parser
from config import globalparser
args = vars(globalparser.getparser().parse_args())
from config import globalconfig
globalconfig.run(args)
print('Train {} with {}.(Running on {})'.format(os.environ['savepath'],
os.environ['datapath'],
os.environ['device']))
import importlib
# Essential network building blocks.
model = importlib.import_module('Loader.Model.' + args['model'])
transform = importlib.import_module('Loader.Transform.train_aug')
# Data loader.
from DataUtils import ImgFolder_5fold as data
# Official packages.
import torch.nn as nn
import torch.optim as optim
# 下面开始进行主干内容
from tools import datainfo
info = datainfo.getdatainfo(os.environ['datapath'])
models, params, modelinfo = model.load(info, args['continue'])
# Global environment setup. import os from config import globalconfig globalconfig.run() globalconfig.update_filename(__file__) # Essential network building blocks. from Networks.Nets import TwoLayerFC from Networks.Nets import ThreeLayerConvNet from Networks.Blocks import LinearReLU # Data loader. from DataUtils import cifar10 # Useful tools. from tools import train_and_check as mtool # Official packages. import torch import torch.nn as nn import torch.optim as optim # Training setup. os.environ['print_every'] = '10' os.environ['save_every'] = '1' TRAIN_EPOCHS=20 LEARNING_RATE=0.1 # GOT DATA train_dataloader, val_dataloader, test_dataloader, sample = cifar10.getdata()