torch加载optimizer时报错：KeyError: ‘optimizer‘

报错

Traceback (most recent call last):
  File "E:/Program Files/PyCharm 2019.2/machinelearning/online/wind_speed/online.py", line 724, in <module>
    un_update_w()
  File "E:/Program Files/PyCharm 2019.2/machinelearning/online/wind_speed/online.py", line 598, in un_update_w
    optimizer.load_state_dict(torch.load(LSTM_PATH)['optimizer'])
KeyError: 'optimizer'

原因
模型保存时并没有保存优化器optimizer参数。

torch.save(model.state_dict(), LSTM_PATH)

解决
训练时保存优化器参数：

state = {'model':model.state_dict(), 'optimizer':optimizer.state_dict(), 'epoch':epoch}
torch.save(state, LSTM_PATH)

然后：

optimizer = torch.optim.Adam(lstm.parameters(), lr=0.001)
checkpoint = torch.load(LSTM_PATH)
optimizer.load_state_dict(checkpoint['optimizer'])