[GPU]: Regression in ONNX models: 'tensor.expand_shape' op mismatch in dynamic dims in output_shape and static_output_shape: static_output_shape has 2 dynamic dims while output_shape has 1 values

[pdhirajkumarprasad]

What happened?

For the given MLIR

module {

  func.func @main_graph( %arg0: !torch.vtensor<[?,12,?,64],f32>, %arg1: !torch.vtensor<[?,12,64,?],f32>) -> !torch.vtensor<[?,12,?,?],f32>  attributes {torch.onnx_meta.ir_version = 7 : si64, torch.onnx_meta.opset_version = 21 : si64, torch.onnx_meta.producer_name = "pytorch", torch.onnx_meta.producer_version = "2.4.0"} {

    %1 = torch.operator "onnx.MatMul"(%arg0, %arg1) : (!torch.vtensor<[?,12,?,64],f32>, !torch.vtensor<[?,12,64,?],f32>) -> !torch.vtensor<[?,12,?,?],f32> 

    return %1 : !torch.vtensor<[?,12,?,?],f32>

  }

}

getting error as

model.torch_onnx.mlir:3:10: error: failed to configure executables
    %1 = torch.operator "onnx.MatMul"(%arg0, %arg1) : (!torch.vtensor<[?,12,?,64],f32>, !torch.vtensor<[?,12,64,?],f32>) -> !torch.vtensor<[?,12,?,?],f32> 
         ^
model.torch_onnx.mlir:3:10: note: see current operation: 

command:

iree-compile --iree-hal-target-backends=rocm --iree-hip-target=gfx942 model.torch_onnx.mlir -o abc.vmfb

Version:

IREE compiler version 3.0.0rc20241114 @ eef2c3a

Steps to reproduce your issue

No response

What component(s) does this issue relate to?

Compiler

Version information

No response

Additional context

No response

[MaheshRavishankar]

#19164 is temporary work-around for this issue.