Activating the cuda feature

[trago]

Hi, as in other comments I must to say this is a great crate for using the onnruntime from rust. I have set it for a little project for inference and it just work well.

But, I'm not sure that I'm actually using cuda for the inference processing. I downloaded the onnxruntime with cuda compiled and use the ORT_STRATEGY=system and set the ORT_LIB_LOCATION, also I set the LD_LIBRARY_PATH so the onnxlibrary is loaded at runtime.

I check the cuda availability using ExecutionProvider::cuda().is_available() and it returns true so I use the cuda provider when creating the session. However I see the same performance as using CPU.

So my question is, how can I setup the cuda feature as in the readme it says that to enable it we must to set this feature. Do you have an example of how to set this feature so I can test?

Thanks

[decahedron1]

Did you enable the cuda Cargo feature?

ort = { version = "1.14.0-alpha.0", features = [ "cuda" ] }

[trago]

That is what probably I'm doing wrong so my question is how to enable it. I added this to my cargo

ort = { git = "https://github.com/pykeio/ort", tag = "v1.14.0-alpha.0", fearures = ["cuda"] } # "1.14.0-alpha.0"

But I have this warning when running the cargo build

warning: /some/path/Cargo.toml: unused manifest key: dependencies.ort.fearures

So I might be doing it wrong. Do you have an example of how to enable cuda?

[decahedron1]

features is spelled wrong

[trago]

you are right, that is funny. Let me try again.

I'll let you know

[trago]

Hey!! Now it is using cuda.

I feel bad for this question hehe, but thanks for letting me see this error from me.

THanks!!