Skip to main content

EagerTensorLinalgExt

tenferro_linalg

Trait EagerTensorLinalgExt 

Source 
pub trait EagerTensorLinalgExt {
    // Required method
    fn solve(
        &self,
        engine: &mut dyn ExecutionEngine,
        b: &EagerTensor,
    ) -> Result<EagerTensor>;
}
Expand description

Tensor-owned linear solve. Other eager linear-algebra operations use EagerSessionLinalgExt inside tenferro_ad::EagerRuntime::with_eager_session.

The tensor-owned entry preserves calling-thread no_grad behavior for tracked solves; it must not be called from inside a borrowed session.

§Examples

let x = a.solve(&mut tenferro_cpu::CpuEngine::serial(), &b)?;
assert_eq!(x.shape(), &[1, 1]);

Required Methods§

Source

fn solve( &self, engine: &mut dyn ExecutionEngine, b: &EagerTensor, ) -> Result<EagerTensor>

The CUDA provider does not inspect the cuSOLVER getrf status flag on the forward solve path, so it performs no per-call non-singularity check and issues no host round trip. An exactly singular input is left to the caller’s verification, as in torch.linalg.solve. CPU providers may still report their own exactly singular factor as Error::Singular.

§Errors

Returns Error::Validation for incompatible matrix, batch, or dtype metadata, Error::Extension for an unsupported dtype or singular system, and Error::RuntimeState when the backend is unavailable.

§Examples
let x = a.solve(&mut tenferro_cpu::CpuEngine::serial(), &b)?;
assert_eq!(x.value()?.as_slice::<f64>()?, &[2.0, 2.0]);

Dyn Compatibility§

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors§