Crate tenferro_xla 

Experimental StableHLO lowering and runtime PJRT plugin loading for tenferro.

This crate is an optional peer executor over tenferro-runtime GraphProgram values. It does not implement TensorBackend and it does not change native CPU, CUDA, or WebGPU execution.

Examples

use tenferro_runtime::{GraphCompiler, TracedTensor};
use tenferro_xla::lower_to_stablehlo;

let x = TracedTensor::from_vec_col_major(vec![2], vec![1.0_f64, 2.0]).unwrap();
let y = (&x + &x).unwrap();
let mut compiler = GraphCompiler::new();
let program = compiler.compile(&y).unwrap();
let module = lower_to_stablehlo(&program).unwrap();
assert!(module.as_str().contains("stablehlo.add"));

Structs

PjrtPlugin

Dynamically loaded PJRT plugin.

StableHloModule

StableHLO MLIR module text plus a deterministic fingerprint.

StableHloModuleFingerprint

SHA-256 fingerprint of StableHLO module text.

XlaExecutor

Experimental peer executor for XLA/PJRT.

XlaExecutorOptions

Options for the experimental XLA executor.

Enums

Error

Error type for StableHLO lowering and runtime PJRT plugin loading.

Constants

TENFERRO_PJRT_GPU_PLUGIN_ENV

Environment variable used for a GPU-specific PJRT plugin path.

TENFERRO_PJRT_PLUGIN_ENV

Environment variable used for the default PJRT plugin path.

Functions

lower_to_stablehlo

Lower a static-shaped graph program to StableHLO MLIR text.

Type Aliases

Result

Result alias for tenferro-xla.