分享个简单的启动代码。
'''
CHANGES:
- imagenet cnns: resnet: http://pytorch.org/docs/master/torchvision/models.html
- places 365 cnns: resnet 18, 50: https://github.com/CSAILVision/places365
- top3 accuracy: https://github.com/pytorch/examples/blob/master/imagenet/main.py
- 训练-验证流程: http://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html#
TODO:
- (训练-验证)-(测试)总流程，代码模块化: https://zhuanlan.zhihu.com/p/29024978
- places: densenet 161
- 测试其他imagenet cnn
- 数据增强，各种套路逐一实现===========================================
- mxnet resnet 152? https://github.com/YanWang2014/iNaturalist
- tf inception-resnet v2？ http://blog.csdn.net/wayne2019/article/details/78210172
'''
#pkill -9 python
#nvidia-smi
import os
import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
from torch.autograd import Variable
from PIL import Image
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
import time
import json
'''
load pretrained model
'''
from functools import partial
import pickle
pickle.load = partial(pickle.load, encoding="latin1")
pickle.Unpickler = partial(pickle.Unpickler, encoding="latin1")
#model = torch.load(model_file, map_location=lambda storage, loc: storage, pickle_module=pickle)
# th architecture to use
arch = 'resnet18_places365' # AlexNet, ResNet18, ResNet50, DenseNet161
model_weight = 'whole_%s.pth.tar' % arch
use_gpu = 1
if use_gpu == 1:
model_conv = torch.load(model_weight, pickle_module=pickle)
else:
model_conv = torch.load(model_weight, map_location=lambda storage, loc: storage, pickle_module=pickle) # model trained in GPU could be deployed in CPU machine like this!
'''
load and transform data
'''
with open('../ai_challenger_scene_train_20170904/scene_train_annotations_20170904.json', 'r') as f: #label文件
label_raw_train = json.load(f)
with open(
