Upgrade to 1.16 - Githubissues

johns10 commented 5 months ago

I'd like to use this with the pretrained pyannote, but it uses ONNX ML opset version 4, which isn't supported by 1.14.

I attempted the upgrade naively and got:

   Compiling ortex v0.1.0 (/Users/johndavenport/Documents/github/desktop-example-app-main/deps/ortex/native/ortex)
error[E0432]: unresolved import `ort::tensor::InputTensor`
  --> src/model.rs:19:5
   |
19 |     tensor::InputTensor, LoggingLevel, OrtError,
   |     ^^^^^^^^^^^^^^^^^^^ no `InputTensor` in `tensor`

error[E0432]: unresolved imports `ort::tensor::DynOrtTensor`, `ort::tensor::FromArray`, `ort::tensor::InputTensor`
 --> src/tensor.rs:4:19
  |
4 | use ort::tensor::{DynOrtTensor, FromArray, InputTensor, TensorElementDataType};
  |                   ^^^^^^^^^^^^  ^^^^^^^^^  ^^^^^^^^^^^ no `InputTensor` in `tensor`
  |                   |             |
  |                   |             no `FromArray` in `tensor`
  |                   no `DynOrtTensor` in `tensor`

For more information about this error, try `rustc --explain E0432`.
error: could not compile `ortex` (lib) due to 2 previous errors

Is this trivial for you to do?

johns10 commented 5 months ago

Happy to make up a pr for this with some guidance.

davenforce commented 5 months ago

There's already a pr for this.

mortont commented 1 week ago

We've updated to 1.19 in main, feel free to open another issue if this doesn't solve your opset version issues

elixir-nx / ortex

Upgrade to 1.16 #36