Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add test onnx.Cast bf16 to f16 #430

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Conversation

AmosLewis
Copy link
Collaborator

@AmosLewis AmosLewis commented Jan 17, 2025

To test llvm/torch-mlir#3963

construct_inputs.log
Failed test at stage construct_inputs with exception:
Numpy doesn't support bfloat16. Please consider modifying the boundary types.

But if it pass manually run

iree-run-module --device=local-task --module=compiled_model.vmfb --function=main --input="1xbf16=1.0"
EXEC @main
result[0]: hal.buffer_view
1xf16=1
module {
  func.func @main(%arg0: !torch.vtensor<[1],bf16>) -> !torch.vtensor<[1],f16> attributes {torch.onnx_meta.ir_version = 10 : si64, torch.onnx_meta.opset_version = 21 : si64, torch.onnx_meta.producer_name = "", torch.onnx_meta.producer_version = ""} {
    %none = torch.constant.none
    %0 = torch.operator "onnx.Cast"(%arg0) {torch.onnx.to = 10 : si64} : (!torch.vtensor<[1],bf16>) -> !torch.vtensor<[1],f16> 
    return %0 : !torch.vtensor<[1],f16>
  }
}
#map = affine_map<(d0) -> (0)>
#map1 = affine_map<(d0) -> (d0)>
module {
  func.func @main(%arg0: tensor<1xbf16>) -> tensor<1xf16> {
    %0 = tensor.empty() : tensor<1xf16>
    %1 = linalg.generic {indexing_maps = [#map, #map1], iterator_types = ["parallel"]} ins(%arg0 : tensor<1xbf16>) outs(%0 : tensor<1xf16>) {
    ^bb0(%in: bf16, %out: f16):
      %2 = arith.extf %in : bf16 to f32
      %3 = arith.truncf %2 : f32 to f16
      linalg.yield %3 : f16
    } -> tensor<1xf16>
    return %1 : tensor<1xf16>
  }
}

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant