[CUDA] cudaErrorInvalidConfiguration with Conv2DTranspose

[jwnhy]

bug描述 Describe the Bug

import paddle as pdl

model = pdl.nn.Conv2DTranspose(47, 47, kernel_size=[46, 46], padding=[46, 15], groups=47, data_format="NCHW")
tensor = pdl.rand([1, 47, 47, 47])
model(tensor)

其他补充信息 Additional Supplementary Information

========= COMPUTE-SANITIZER                                                                                                         16:23:53 [83/1230]
W0218 16:23:53.068138 1700471 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 9.0, Driver API Version: 12.8, Runtime API Version
: 12.3
W0218 16:23:53.069804 1700471 gpu_resources.cc:164] device: 0, cuDNN Version: 9.7.
========= Program hit cudaErrorInvalidConfiguration (error 9) due to "invalid configuration argument" on CUDA API call to cudaLaunchKernel.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame: [0x4aa4a5]
=========                in /lib/x86_64-linux-gnu/libcuda.so.1
=========     Host Frame:cudaLaunchKernel [0x7466d]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/base/../../../../libcudart.so.12
=========     Host Frame:paddle::operators::math::DepthwiseConvInputGradFunctor<phi::GPUContext, float, false>::operator()(phi::GPUContext const&, phi
::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator
<int> > const&, std::vector<int, std::allocator<int> > const&, phi::DenseTensor*, common::DataLayout) [0x156ce8c]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/base/../libs/libphi_kernel_gpu.so
=========     Host Frame:void phi::DepthwiseConv2dTransposeKernel<float, phi::GPUContext>(phi::GPUContext const&, phi::DenseTensor const&, phi::DenseT
ensor const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > con
st&, paddle::experimental::IntArrayBase<phi::DenseTensor> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > cons
t&, int, std::vector<int, std::allocator<int> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, phi::D
enseTensor*) [0x15d1470]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/base/../libs/libphi_kernel_gpu.so
=========     Host Frame:paddle::experimental::depthwise_conv2d_transpose(paddle::Tensor const&, paddle::Tensor const&, std::vector<int, std::allocato
r<int> > const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, paddle::experimental::IntArrayBase<padd
le::Tensor> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int, std::vector<int, std::allocator<int>
> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) [0x281bf23]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/base/../libs/libphi.so
=========     Host Frame:depthwise_conv2d_transpose_ad_func(paddle::Tensor const&, paddle::Tensor const&, std::vector<int, std::allocator<int> >, std:
:vector<int, std::allocator<int> >, std::vector<int, std::allocator<int> >, paddle::experimental::IntArrayBase<paddle::Tensor>, std::__cxx11::basic_st
ring<char, std::char_traits<char>, std::allocator<char> >, int, std::vector<int, std::allocator<int> >, std::__cxx11::basic_string<char, std::char_tra
its<char>, std::allocator<char> >) [0x6b0cea5]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/base/libpaddle.so
=========     Host Frame:paddle::pybind::eager_api_depthwise_conv2d_transpose(_object*, _object*, _object*) [0x4c37b0a]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/base/libpaddle.so
=========     Host Frame:cfunction_call in /usr/local/src/conda/python-3.12.9/Objects/methodobject.c:537 [0x14d2d3]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/bin/python3
=========     Host Frame:_PyObject_MakeTpCall in /usr/local/src/conda/python-3.12.9/Objects/call.c:240 [0x11e38a]
=========                in /opt/ext/jwnhy/miniconda3/envs/pdl300/bin/python3

W0218 16:26:02.785025 1701551 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 9.0, Driver API Version: 12.8, Runtime API Version: 12.3
W0218 16:26:02.785495 1701551 gpu_resources.cc:164] device: 0, cuDNN Version: 9.7.
Traceback (most recent call last):
  File "/home/jwnhy/gpu_fuzz/poc/poc13.py", line 5, in <module>
    model(tensor)
  File "/opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/nn/layer/layers.py", line 1426, in __call__
    return self.forward(*inputs, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/nn/layer/conv.py", line 883, in forward
    out = F.conv2d_transpose(
          ^^^^^^^^^^^^^^^^^^^
  File "/opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/nn/functional/conv.py", line 1275, in conv2d_transpose
    return _add_with_axis(pre_bias, bias, axis=channel_dim)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/tensor/math.py", line 1152, in _add_with_axis
    return _elementwise_op_with_axis(x, y, axis, name, "add")
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/ext/jwnhy/miniconda3/envs/pdl300/lib/python3.12/site-packages/paddle/tensor/math.py", line 1146, in _elementwise_op_with_axis
    return op(x, y)
           ^^^^^^^^
RuntimeError: (PreconditionNotMet) The meta data must be valid when call the mutable data function.
  [Hint: Expected valid() == true, but received valid():0 != true:1.] (at ../paddle/phi/core/dense_tensor.cc:111)

[LiYuRio]

api参数值设置不合理，导致在计算过程中得到了0维tensor，导致报错