Skip to content

Instantly share code, notes, and snippets.

@AmosLewis

AmosLewis/NumToTensor.mlir

Created January 16, 2023 07:26
Show Gist options
  • Save AmosLewis/fca6b0d16ee325fcf7ee400459f4fd40 to your computer and use it in GitHub Desktop.
Save AmosLewis/fca6b0d16ee325fcf7ee400459f4fd40 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
NumToTensor.mlir
func.func @torch.prim.NumToTensor.Scalar() -> !torch.vtensor<[],f64> {
%float8.000000e00 = torch.constant.float 8.000000e+00
%1 = "torch.prim.NumToTensor.Scalar"(%float8.000000e00) : (!torch.float) -> !torch.vtensor<[],f64>
return %1 : !torch.vtensor<[],f64>
}
@AmosLewis
Copy link
Author

For comparison with F64: F32 successfuly

func.func @torch.prim.NumToTensor.Scalar() -> !torch.vtensor<[],f32> {
  %float8.000000e00 = torch.constant.float 8.000000e+00
  %1 = "torch.prim.NumToTensor.Scalar"(%float8.000000e00) : (!torch.float) -> !torch.vtensor<[],f32>
  return %1 : !torch.vtensor<[],f32>
}

--->

module {
  func.func @torch.prim.NumToTensor.Scalar() -> !torch.vtensor<[],f32> {
    %float8.000000e00 = torch.constant.float 8.000000e+00
    %0 = "tosa.const"() {value = dense<8.000000e+00> : tensor<f32>} : () -> tensor<f32>
    %1 = torch_c.from_builtin_tensor %0 : tensor<f32> -> !torch.vtensor<[],f32>
    return %1 : !torch.vtensor<[],f32>
  }
}

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment