error: 'tensor.cast' op operand type 'tensor<0xi64>' and result type 'tensor<12xi64>' are cast incompatible

[silvasean]

@vid-999 this appears to be a bug in #1066 PTAL

Reproduce with

torch-mlir-opt -convert-torch-to-linalg repro.mlir 

module {
  func.func @forward(%arg0: !torch.vtensor<[24],si64>, %arg1: !torch.vtensor<[4],si64>, %arg2: !torch.vtensor<[12],si64>, %arg3: !torch.vtensor<[10,4],f32>) -> !torch.vtensor<[4,4],f32> {
    %int0 = torch.constant.int 0
    %int-1 = torch.constant.int -1
    %false = torch.constant.bool false
    %none = torch.constant.none
    %result0, %result1, %result2, %result3 = torch.aten.embedding_bag.padding_idx %arg3, %arg2, %arg1, %false, %int0, %false, %none, %false, %int-1 : !torch.vtensor<[10,4],f32>, !torch.vtensor<[12],si64>, !torch.vtensor<[4],si64>, !torch.bool, !torch.int, !torch.bool, !torch.none, !torch.bool, !torch.int -> !torch.vtensor<[4,4],f32>, !torch.vtensor<[12],si64>, !torch.vtensor<[4],si64>, !torch.vtensor<[4],si64>
    return %result0 : !torch.vtensor<[4,4],f32>
  }
}

Full error:

/tmp/repro4.mlir:7:46: error: 'tensor.cast' op operand type 'tensor<0xi64>' and result type 'tensor<12xi64>' are cast incompatible
    %result0, %result1, %result2, %result3 = torch.aten.embedding_bag.padding_idx %arg3, %arg2, %arg1, %false, %int0, %false, %none, %false, %int-1 : !torch.vtensor<[10,4],f32>, !torch.vtensor<[12],si64>, !torch.vtensor<[4],si64>, !torch.bool, !torch.int, !torch.bool, !torch.none, !torch.bool, !torch.int -> !torch.vtensor<[4,4],f32>, !torch.vtensor<[12],si64>, !torch.vtensor<[4],si64>, !torch.vtensor<[4],si64>
                                             ^
/tmp/repro4.mlir:7:46: note: see current operation: %22 = "tensor.cast"(%21) : (tensor<0xi64>) -> tensor<12xi64>