个人报错问题 持续总结

代码框内为报错

代码框下是解释

---

  File "C:\Users\52490\AppData\Local\Programs\Python\Python37\lib\urllib\request.py", line 1352, in do_open
    raise URLError(err)

ConnectionResetError: [WinError 10054] Զ▒▒▒▒▒▒ǿ▒ȹر▒▒▒һ▒▒▒▒▒е▒▒▒▒ӡ▒

网络问题

---

AttributeError: Can't pickle local object 'checkpoint.begin_background..bg_target'

（目前理解）无法将本地对象腌制成二进制格式进行存储，解决方式是把这个函数抽出来作为静态函数，代码如下

# 原代码
@staticmethod
def begin_background(self):
    self.queue = Queue()

    def bg_target(queue):
        while True:
            if not queue.empty():
                filename, tensor = queue.get()
                if filename is None: break
                imageio.imwrite(filename, tensor.numpy())

        self.process = [
            Process(target=bg_target, args=(self.queue,)) \
            for _ in range(self.n_processes)
        ]
        
        for p in self.process: p.start()

# 修改后的代码
    @staticmethod
    def bg_target(queue):
        while True:
            if not queue.empty():
                filename, tensor = queue.get()
                if filename is None: break
                imageio.imwrite(filename, tensor.numpy())

    def begin_background(self):
        self.queue = Queue()

        self.process = [
            Process(target=self.bg_target, args=(self.queue,)) \
            for _ in range(self.n_processes)
        ]
        
        for p in self.process: p.start()

 

---

torch torchvision升版本

pytorch与torchvision的更新，卸载和秒安装（看这一篇就够了） - 灰信网（软件开发博客聚合）

---

Epoch 0: StepDecay set learning rate to 0.05000000074505806.
Traceback (most recent call last):
  File "/mnt/zyh/projects/srgan/train.py", line 212, in <module>
    train()
  File "/mnt/zyh/projects/srgan/train.py", line 147, in train
    loss = trainforinit(lr_patch, hr_patch)
  File "/home/zyh/.conda/envs/torchtf/lib/python3.9/site-packages/tensorlayerx/model/core.py", line 564, in __call__
    loss = self.net_with_train(data, label)
  File "/home/zyh/.conda/envs/torchtf/lib/python3.9/site-packages/tensorlayerx/model/utils.py", line 196, in __call__
    self.optimizer.apply_gradients(zip(grads, self.train_weights))
  File "/home/zyh/.conda/envs/torchtf/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "/home/zyh/.conda/envs/torchtf/lib/python3.9/site-packages/tensorlayerx/optimizers/torch_optimizers.py", line 627, in apply_gradients
    F.sgd(params_with_grad,
TypeError: sgd() missing 1 required keyword-only argument: 'maximize'

解决方案：找到报错地点，加上maximize=False

F.sgd(params_with_grad,
                  d_p_list,
                  momentum_buffer_list,
                  weight_decay=weight_decay,
                  momentum=momentum,
                  lr=lr,
                  dampening=dampening,
                  nesterov=nesterov,
                  maximize=False)