CPU 环境反向报错：While running Node (GroupNormGradNode) raises an EnforceNotMet exception

[wangguan1995]

bug描述 Describe the Bug

W0418 04:25:20.342784 80253 backward.cc:437] While running Node (GroupNormGradNode) raises an EnforceNotMet exception
Traceback (most recent call last):
File "/workspace/MathorCup/PaddleScience/competition/MathorCup2025/main.py", line 361, in
train(config_p)
File "/workspace/MathorCup/PaddleScience/competition/MathorCup2025/main.py", line 299, in train
loss.backward()
File "/usr/local/lib/python3.10/dist-packages/decorator.py", line 232, in fun
return caller(func, *(extras + args), **kw)
File "/usr/local/lib/python3.10/dist-packages/paddle/base/wrapped_decorator.py", line 40, in impl
return wrapped_func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/paddle/base/framework.py", line 704, in impl
return func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/paddle/base/dygraph/tensor_patch_methods.py", line 358, in backward
core.eager.run_backward([self], grad_tensor, retain_graph)
ValueError: (InvalidArgument) Required tensor shall not be nullptr, but received nullptr.
[Hint: tensor should not be null.] (at ../paddle/phi/core/device_context.cc:141)

其他补充信息 Additional Supplementary Information

No response

[HydrogenSulfate]

#72358