Add dtype check for MatmulInferMeta

[zrr1999]

PR Category

Operator Mechanism

PR Types

Bug fixes

Description

linear 精度报错与 matmul 相关，暴露出 matmul 的报错信息不明确的问题，给MatmulInferMeta添加额外的检查可以让报错信息更明确。

matmul 的 x 和 y 类型不一致时，目前出现过两种报错
第一种报错
(InvalidArgument) The type of data we are trying to retrieve (float32) does not match the type of data (float16) currently contained in the container.
[Hint: Expected dtype() == phi::CppTypeToDataType::Type(), but received dtype():15 != phi::CppTypeToDataType::Type():10.] (at ../paddle/phi/core/dense_tensor.cc:160)
[operator < linear > error]
第二种报错
Traceback (most recent call last):
File "/workspace/PaddleAPITest/tester/accuracy.py", line 92, in test
exec(code.core_compiled, exec_globals, exec_locals)
File "", line 10, in
File "/workspace/PaddleAPITest/.venv/lib/python3.10/site-packages/torch/utils/_device.py", line 104, in torch_function
return func(*args, **kwargs)
RuntimeError: expected mat1 and mat2 to have the same dtype, but got: c10::Half != float
paddle kernel的核心源码如下

可以看到实际上是要求x和y输入类型一致的，但是不一致时并不会直接报错，因为 MatmulInferMeta 里并没有检查，所以需要添加检查从而减少后续错误排查的工作量。

Pcard-67164

[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.