expressions.rs

extern crate rand;
extern crate tensorflow;

use std::error::Error;
use std::result::Result;
use tensorflow::expr::{Compiler, Placeholder};
use tensorflow::Code;
use tensorflow::Graph;
use tensorflow::Session;
use tensorflow::SessionOptions;
use tensorflow::SessionRunArgs;
use tensorflow::Status;
use tensorflow::Tensor;

#[cfg_attr(feature = "examples_system_alloc", global_allocator)]
#[cfg(feature = "examples_system_alloc")]
static ALLOCATOR: std::alloc::System = std::alloc::System;

struct Checker {
    success: bool,
    epsilon: f32,
}

impl Checker {
    fn new(epsilon: f32) -> Self {
        Checker {
            success: true,
            epsilon: epsilon,
        }
    }

    fn check(&mut self, name: &str, expected: f32, actual: f32) {
        let success = (expected - actual).abs() < self.epsilon;
        println!(
            "Checking {}: expected {}, got {}. {}",
            name,
            expected,
            actual,
            if success { "Success!" } else { "FAIL" }
        );
        self.success &= success;
    }

    fn result(&self) -> Result<(), Box<dyn Error>> {
        if self.success {
            Ok(())
        } else {
            Err(Box::new(Status::new_set(
                Code::Internal,
                "At least one check failed",
            )?))
        }
    }
}

fn main() -> Result<(), Box<dyn Error>> {
    // Build the graph
    let mut g = Graph::new();
    let y_node = {
        let mut compiler = Compiler::new(&mut g);
        let x_expr = <Placeholder<f32>>::new_expr(&vec![2], "x");
        compiler.compile(x_expr * 2.0f32 + 1.0f32)?
    };
    let x_node = g.operation_by_name_required("x")?;
    // This is another valid way to get x_node and y_node:
    // let (x_node, y_node) = {
    //   let mut compiler = Compiler::new(&mut g);
    //   let x_expr = <Placeholder<f32>>::new_expr(&vec![2], "x");
    //   let x_node = compiler.compile(x_expr.clone())?;
    //   let y_node = compiler.compile(x_expr * 2.0f32 + 1.0f32)?;
    //   (x_node, y_node)
    // };
    let options = SessionOptions::new();
    let session = Session::new(&options, &g)?;

    // Evaluate the graph.
    let mut x = <Tensor<f32>>::new(&[2]);
    x[0] = 2.0;
    x[1] = 3.0;
    let mut step = SessionRunArgs::new();
    step.add_feed(&x_node, 0, &x);
    let output_token = step.request_fetch(&y_node, 0);
    session.run(&mut step).unwrap();

    // Check our results.
    let output_tensor = step.fetch::<f32>(output_token)?;
    let mut checker = Checker::new(1e-3);
    checker.check("output_tensor[0]", 5.0, output_tensor[0]);
    checker.check("output_tensor[1]", 7.0, output_tensor[1]);
    checker.result()
}