paddle error
RuntimeError: (NotFound) No Output(X@GRAD) found for BatchNormGrad operator.
[Hint: Expected ctx->HasOutput(framework::GradVarName(“X”)) == true, but received ctx->HasOutput(framework::GradVarName(“X”)):0 != true:1.] (at /paddle/paddle/fluid/operators/batch_norm_op.cc:468)

The problem of BN layer
Modification method: Precede the BN layer with

if self.data_bn:
    x.stop_gradient = False
x = self.data_bn(x)

Similar Posts:

• Python: why use it optimizer.zero_ grad()
• PyTorch :TypeError: exceptions must derive from BaseException
• Solutions to errors encountered by Python
• xxx/labelKeypoint/utils/qt.py:81: RuntimeWarning: invalid value encountered in double_scalars
• [Solved] Pytorch: loss.backward (retain_graph = true) of back propagation error
• [Solved] Operator overload must take either zero or one argument error
• [Solved] An error occurred when paddlepaddle iterated data: typeerror: ‘function’ object is not iterative
• Tensorflowcenter {typeerror} non hashable type: “numpy. Ndarray”
• Solution to GPU memory leak problem of tensorflow operation efficiency
• Tensorflow gradients TypeError: Fetch argument None has invalid type