tenferro-ad 0.3.0

Eager runtime, eager tensors, and traced AD extension traits for tenferro.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81

use std::sync::Arc;

use tenferro_ad::{EagerRuntime, EagerTensor};
use tenferro_cpu::CpuBackend;
#[cfg(feature = "cuda")]
use tenferro_gpu::cuda::CudaBackend;
#[cfg(feature = "webgpu")]
use tenferro_gpu::webgpu::WebGpuBackend;
use tenferro_runtime::Tensor;

#[test]
fn eager_runtime_constructors_return_typed_results() {
    let _: fn() -> tenferro_ad::Result<Arc<EagerRuntime>> = EagerRuntime::new;
    let _: fn(CpuBackend) -> tenferro_ad::Result<Arc<EagerRuntime>> =
        EagerRuntime::with_cpu_backend;
    let _: fn(CpuBackend, &tenferro_ad::AdContext) -> tenferro_ad::Result<Arc<EagerRuntime>> =
        EagerRuntime::with_cpu_backend_and_ad_context;

    #[cfg(feature = "cuda")]
    {
        let _: fn(CudaBackend) -> tenferro_ad::Result<Arc<EagerRuntime>> =
            EagerRuntime::with_cuda_backend;
        let _: fn(CudaBackend, &tenferro_ad::AdContext) -> tenferro_ad::Result<Arc<EagerRuntime>> =
            EagerRuntime::with_cuda_backend_and_ad_context;
    }

    #[cfg(feature = "webgpu")]
    {
        let _: fn(WebGpuBackend) -> tenferro_ad::Result<Arc<EagerRuntime>> =
            EagerRuntime::with_webgpu_backend;
        let _: fn(
            WebGpuBackend,
            &tenferro_ad::AdContext,
        ) -> tenferro_ad::Result<Arc<EagerRuntime>> =
            EagerRuntime::with_webgpu_backend_and_ad_context;
    }
}

#[test]
fn eager_runtime_replaces_eager_context_public_name() {
    let runtime = EagerRuntime::with_cpu_backend(CpuBackend::new()).unwrap();
    let x = EagerTensor::requires_grad_in(
        Tensor::from_vec_col_major(vec![2], vec![1.0_f64, 2.0]).unwrap(),
        runtime.clone(),
    )
    .unwrap();
    let loss = x.mul(&x).unwrap().reduce_sum(Some(&[0])).unwrap();

    loss.backward().unwrap();

    assert_eq!(
        x.grad().unwrap().unwrap().as_slice::<f64>().unwrap(),
        &[2.0, 4.0]
    );
    runtime.clear_grads().unwrap();
    assert!(x.grad().unwrap().is_none());
}

#[test]
fn eager_runtime_synchronize_is_available_and_cpu_noop() {
    let runtime = EagerRuntime::with_cpu_backend(CpuBackend::new()).unwrap();

    runtime.synchronize().unwrap();
}

#[test]
fn eager_runtime_grad_accumulation_keeps_slot_locked_through_update() {
    let source = include_str!("../../src/eager.rs");

    assert!(
        !source.contains("let mut staged = Vec::new();"),
        "gradient accumulation must not stage slot updates after releasing the per-slot lock"
    );
}

#[cfg(feature = "webgpu")]
#[test]
fn eager_runtime_accepts_webgpu_backend_constructor() {
    let _ctor: fn(WebGpuBackend) -> tenferro_ad::Result<std::sync::Arc<EagerRuntime>> =
        EagerRuntime::with_webgpu_backend;
}