TypeError: Unsupported type <class'NoneType'> error when creating autoencoder in Chainer

I'm trying to make a self-encoder with Python using a Chainer.I wrote and executed the following code:

class Autoencoder (Chain):
    def__init__(self):
       super().__init__()
       with self.init_scope():
           self.l1 = L. Linear (3,2)
           self.l2 = L. Linear(2,3)
    def__call__(self, x):
       h1 = self.l1(x)
       h2 = self.l2(h1) 

       return 2

class Dataset(dataset.DatasetMixin):
    def_init__(self, number_of_data, show_initial=False):

       noise_level = 1

       self.data=np.zeros(number_of_data,3), dtype=np.float32)

       OA_vector=np.array ([3,2,1])
       OB_vector=np.array ([2,-1,1])

       t=np.random.uniform (-0.5, 0.5, number_of_data)
       s=np.random.uniform (-0.5, 0.5, number_of_data)

       for i in range (0, number_of_data):
           noise=np.random.uniform (-noise_level, noise_level, 3)
           self.data[i]=t[i]*OA_vector+s[i]*OB_vector+noise

   def__len__(self):
        return self.data.shape [0]

   def get_example(self,idx):
       return self.data [idx]

if__name__=="__main__":

    n_epoch = 5
    batch_size=100

    number_of_data = 1000# Number of data
    train_data = Dataset(number_of_data, False)

    model=Autoencoder()

    optimizer=optimizers.SGD(lr=0.05).setup(model)
    train_iter=iterators.SerialIterator(train_data,batch_size)

    updater=training.StandardUpdater(train_iter, optimizer, device=0)
    trainer=training.Trainer(updater,(n_epoch, "epoch"), out="result")

    trainer.run()

Dataset(number_of_data, False) is a function that obtains a three-dimensional vector of the number_of_data number.If False is set to True, the data obtained can be visualized.

The following error occurs when running:What is the cause?

Exception in main training loop: Unsupported type <class 'NoneType' >
Traceback (most recent call last):
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/trainer.py", line 308, in run
    update()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 149, in update
    self.update_core()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 164, in update_core
    optimizer.update(loss_func, in_arrays)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/optimizer.py", line 655, in update
    loss.backward(loss_scale=self._loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line966, inbackward
    self._backward_main(retain_grad, loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line 1095, in_backward_main
    target_input_indexes, out_grad, in_grad)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line548, inbackward_accumulate
    gxs=self.backward(target_input_indexes, grad_outputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 73, inbackward
    return ReLUGrad2(y).apply(gy,))
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 258, apply
    outputs=self.forward(in_data)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 368, in forward
    return self.forward_cpu(inputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 97, in forward_cpu
    y = (self.b>0) * inputs [0]
  File "cupy/core/core.pyx", line 1310, incupy.core.core.ndarray.__mul_
  File "cupy/core/elementwise.pxi", line 753, incupy.core.core.ufunc.__call__
  File "cupy/core/elementwise.pxi", line 68, incupy.core.core._preprocess_args
Will finalize trainer extensions and updater before re-raising the exception.
Traceback (most recent call last):
  File "AC.py", line 71, in<module>
    trainer.run()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/trainer.py", line 322, in run
    Six.release (*sys.exc_info())
  File"/home/****/.local/lib/python 3.5/site-packages/six.py", line 693, in erase
    raise value
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/trainer.py", line 308, in run
    update()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 149, in update
    self.update_core()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 164, in update_core
    optimizer.update(loss_func, in_arrays)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/optimizer.py", line 655, in update
    loss.backward(loss_scale=self._loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line966, inbackward
    self._backward_main(retain_grad, loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line 1095, in_backward_main
    target_input_indexes, out_grad, in_grad)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line548, inbackward_accumulate
    gxs=self.backward(target_input_indexes, grad_outputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 73, inbackward
    return ReLUGrad2(y).apply(gy,))
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 258, apply
    outputs=self.forward(in_data)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 368, in forward
    return self.forward_cpu(inputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 97, in forward_cpu
    y = (self.b>0) * inputs [0]
  File "cupy/core/core.pyx", line 1310, incupy.core.core.ndarray.__mul_
  File "cupy/core/elementwise.pxi", line 753, incupy.core.core.ufunc.__call__
  File "cupy/core/elementwise.pxi", line 68, incupy.core.core._preprocess_args
TypeError: Unsupported type <class 'NoneType' >

By the way, if you comment out trainer.run(), you won't get an error (of course, you won't start learning…)

There is a cupy error, so I thought it might be related to GPU.

updater=training.StandardUpdater(train_iter, optimizer, device=-1)

Now that I've tried it,

Exception in main training loop: unsupported operand type(s) for*: 'bool' and 'NoneType'
Traceback (most recent call last):
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/trainer.py", line 308, in run
    update()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 149, in update
    self.update_core()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 164, in update_core
    optimizer.update(loss_func, in_arrays)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/optimizer.py", line 655, in update
    loss.backward(loss_scale=self._loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line966, inbackward
    self._backward_main(retain_grad, loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line 1095, in_backward_main
    target_input_indexes, out_grad, in_grad)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line548, inbackward_accumulate
    gxs=self.backward(target_input_indexes, grad_outputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 73, inbackward
    return ReLUGrad2(y).apply(gy,))
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 258, apply
    outputs=self.forward(in_data)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 368, in forward
    return self.forward_cpu(inputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 97, in forward_cpu
    y = (self.b>0) * inputs [0]
Will finalize trainer extensions and updater before re-raising the exception.
Traceback (most recent call last):
  File "AC.py", line 70, in<module>
    trainer.run()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/trainer.py", line 322, in run
    Six.release (*sys.exc_info())
  File"/home/****/.local/lib/python 3.5/site-packages/six.py", line 693, in erase
    raise value
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/trainer.py", line 308, in run
    update()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 149, in update
    self.update_core()
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/training/updaters/standard_updater.py", line 164, in update_core
    optimizer.update(loss_func, in_arrays)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/optimizer.py", line 655, in update
    loss.backward(loss_scale=self._loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line966, inbackward
    self._backward_main(retain_grad, loss_scale)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/variable.py", line 1095, in_backward_main
    target_input_indexes, out_grad, in_grad)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line548, inbackward_accumulate
    gxs=self.backward(target_input_indexes, grad_outputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 73, inbackward
    return ReLUGrad2(y).apply(gy,))
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 258, apply
    outputs=self.forward(in_data)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/function_node.py", line 368, in forward
    return self.forward_cpu(inputs)
  File"/home/****/.local/lib/python 3.5/site-packages/chainer/functions/activation/relu.py", line 97, in forward_cpu
    y = (self.b>0) * inputs [0]
TypeError: unsupported operand type(s) for*: 'bool' and 'NoneType'

The error appears.

python machine-learning deep-learning chainer

2022-09-30 14:05

1 Answers

Error due to lack of correct answer data and loss function.
If you write without using trainer

for batch training_iter:
    put con=chain.dataset.concat_examples(batch)#dataset[0],dataset[1],... in one array
    put con=chain.Variable(con)#array into chain.Variable
    pred=model(con)# guess
    loss=F.mean_absolute_error(pred,con)#Compare with guesses and correct answer data (in the case of AE itself)
    model.cleargrads()#Initialize previously calculated gradient
    calculate gradient for loss.backward()#loss
    Optimize the model based on optimizer.update()#gradient

Learning can be accomplished through the loop.
If you want to use trainer, use chianer.links.Classifier code (https://docs.chainer.org/en/stable/reference/generated/chainer.links.Classifier.html) to create a new model that connects model and loss function.
For example:

from chain import reporter
class AEWrapper (Chain):

    def_init__(self, predictor,
                 lossfun=F.mean_absolute_error, ):
        super(AEWrapper,self).__init__()
        self.lossfun=lossfun
        with self.init_scope():
            self.predictor=predictor

    def forward (self, *args, **kwargs):
        self.y = None
        self.loss=None

        self.y = self.predictor(*args)
        self.loss=self.lossfun(self.y, *args)
        reporter.report({'loss':self.loss},self)
        return self.loss

By the way, according to J.J. Sakurai's code, both the learning results and the process will disappear when finished, so I think it's better to take logs and print them properly.

n_epoch=5
batch_size=100
number_of_data = 1000# Number of data
train_data = Dataset(number_of_data, False)

model=Autoencoder()
model=AEWrapper(model)

optimizer=chainer.optimizers.SGD(lr=0.05).setup(model)
train_iter=chain.iterators.SerialIterator(train_data,batch_size)

updater=chainer.training.StandardUpdater(train_iter, optimizer, device=0)
trainer=chainer.training.Trainer(updater, (n_epoch, "epoch"), out="result")
trainer.extend(extensions.LogReport(keys=["main/loss"],
                         trigger=training.trigger.IntervalTrigger(1, 'epoch'))))
trainer.extend(extensions.PrintReport(['epoch','main/loss'],
                trigger = training.trigger.IntervalTrigger(1, 'epoch') 
trainer.extend(extensions.snapshot_object(model,
                            'model_{.updater.iteration}.npz'), trigger=(1, 'epoch')
trainer.run()

epoch main/loss 
1           1.31194     
2           1.0665      
3           0.883515    
4           0.743101    
5           0.635571

2022-09-30 14:05

If you have any answers or tips

Popular Tags

python x 4647

android x 1593

java x 1494

javascript x 1427

c x 927

c++ x 878

ruby-on-rails x 696

php x 692

python3 x 685

html x 656