[0-size Tensor No.160、162、164] Add 0-size Tensor support for conv1d_transpose

[co63oc]

PR Category

Execute Infrastructure

PR Types

Improvements

Description

[0-size Tensor No.160、162、164] Add 0-size Tensor support for conv1d_transpose

paddle.nn.functional.conv1d_transpose
paddle.nn.functional.conv2d_transpose
paddle.nn.functional.conv3d_transpose

修改 infermeta ，如果维度为0，不检查output_size 和 infer_shape 比较
符号推导没有对应代码

Paddle/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/binary_infer_sym.cc

Line 447 in 3ad243c

bool ConvTransposeFunction(pir::Operation *op,

kernel 使用到depthwise_conv 和 conv kernel， 修改 depthwise_conv 和 conv kernel
正向测试返回填充时torch为错误，这里按0填充

python engine.py --accuracy=True --api_config='paddle.nn.functional.conv3d(Tensor([2, 1, 1, 1, 1],"float32"
), Tensor([3, 1, 3, 3, 0],"float32"), None, padding=1, stride=1, dilation=1, groups=1, data_format="NCDHW", )'

反向测试filter为填充0

import torch
x=torch.randn([0,1,1])
y=torch.randn([1,1,1])
x.requires_grad=True
y.requires_grad=True
z=torch.nn.functional.conv_transpose1d(x, y, groups=1)
z.sum().backward()
y.grad

tensor([[[0.]]])

PaddleAPITest 测试
GPU测试为torch error

CPU测试为torch error

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Please upload report for BASE (develop@3efb8db). Learn more about missing BASE report.

Additional details and impacted files

@@             Coverage Diff             @@
##             develop    #73698   +/-   ##
===========================================
  Coverage           ?   100.00%           
===========================================
  Files              ?         1           
  Lines              ?         2           
  Branches           ?         0           
===========================================
  Hits               ?         2           
  Misses             ?         0           
  Partials           ?         0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[DanielSun11]

关于grad部分的修改存在一些疑问：
前向过程的输出out shape如下所示：

也就是可能存在x和filter都不为0 size的情况下output的shape为0？
此时grad 的dout应该也为0-size，这种情况下需不需要full一下dx和dfilter？
上述这个case可能比较难找，请co师傅研判下需不需要应对这种情况。

[co63oc]

关于grad部分的修改存在一些疑问： 前向过程的输出out shape如下所示： 也就是可能存在x和filter都不为0 size的情况下output的shape为0？ 此时grad 的dout应该也为0-size，这种情况下需不需要full一下dx和dfilter？ 上述这个case可能比较难找，请co师傅研判下需不需要应对这种情况。

https://github.com/PaddlePaddle/Paddle/blob/1d608281f1187715b80e31e1a3d00d6f09f92c6c/paddle/phi/kernels/impl/conv_transpose_grad_kernel_impl.h#L150C2-L150C3

默认是填充0，如果出错应该是有异常，可以按异常再查找问题，因为出现可能性不大，应该暂时可以不用修改

[DanielSun11]

LGTM