将deep_sort跟踪模型ckpt.t7转为pt格式，用于pc端的推断。
def cvt_model():
print("===> Loading model")
model = Net()
modelname = 'ckpt.t7'
checkpoint = torch.load(savedataroot + modelname)
model.load_state_dict(checkpoint['net_dict']) # 从字典中依次读取，具体值查看字典更改
print('===> Load last checkpoint data')
# 模型转换，Torch Script
model.eval()
example = torch.rand(4,3,128,64)
traced_script_module = torch.jit.trace(model, example)
traced_script_module.save("deep_model.pt")
print("Export of model.pt complete!")
