mil_infer 0.5.2

Sovereign Apple Neural Engine compiler, decompiler, and runtime: native HWX Mach-O emitter + decompiler, MIL DAG graph scheduler, on-chip ANE backprop/SGD training, zero-copy IOSurface/Espresso execution.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
use mil_infer::espresso::EspressoProgram;
use std::path::Path;
fn main() {
    let m =
        EspressoProgram::load_precompiled(Path::new("/tmp/maskadd_o/maskadd.macho"), &[]).unwrap();
    for d in &m.input_descs {
        println!(
            "in {} k{} h{} w{} ps{}",
            d.name, d.k, d.h, d.w, d.plane_stride
        );
    }
    for d in &m.output_descs {
        println!(
            "out {} k{} h{} w{} ps{}",
            d.name, d.k, d.h, d.w, d.plane_stride
        );
    }
}