rumus 0.3.1

A native-Rust deep learning framework with explicit memory safety and hardware acceleration
Documentation 
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
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132

// SPDX-License-Identifier: Apache-2.0 OR MIT
//! Thread-local autograd state: tape, GradId generator, and `no_grad` guard.
//!
//! Each thread owns its own [`Tape`] and id counter — no global locks.
//! This mirrors PyTorch's thread-local dispatch model.

use std::cell::RefCell;

use crate::autograd::Tape;
use crate::tensor::GradId;

// ---------------------------------------------------------------------------
// Thread-local context
// ---------------------------------------------------------------------------

/// Per-thread autograd bookkeeping.
struct AutogradContext {
    /// The active tape.  `None` before the first tracked op or after
    /// `take_tape()` consumes it for backward.
    tape: Option<Tape>,
    /// Monotonically increasing GradId generator.
    next_grad_id: usize,
    /// Nesting depth of `no_grad` guards.  Recording is suppressed while
    /// this is > 0.
    no_grad_depth: usize,
}

thread_local! {
    static CONTEXT: RefCell<AutogradContext> = RefCell::new(AutogradContext {
        tape: None,
        next_grad_id: 0,
        no_grad_depth: 0,
    });
}

// ---------------------------------------------------------------------------
// Public API (crate-internal)
// ---------------------------------------------------------------------------

/// Generate a fresh [`GradId`], unique within this thread.
///
/// The counter is monotonically increasing and never resets.
pub(crate) fn next_grad_id() -> GradId {
    CONTEXT.with(|c| {
        let mut ctx = c.borrow_mut();
        let id = GradId(ctx.next_grad_id);
        ctx.next_grad_id += 1;
        id
    })
}

/// Execute `f` with a mutable reference to the active [`Tape`].
///
/// If no tape exists yet, one is lazily created.  Returns `None` if
/// `no_grad` is active (recording is suppressed).
pub(crate) fn with_tape<F, R>(f: F) -> Option<R>
where
    F: FnOnce(&mut Tape) -> R,
{
    CONTEXT.with(|c| {
        let mut ctx = c.borrow_mut();
        if ctx.no_grad_depth > 0 {
            return None;
        }
        // Lazily create the tape on the first tracked operation.
        if ctx.tape.is_none() {
            ctx.tape = Some(Tape::new());
        }
        Some(f(ctx.tape.as_mut().unwrap()))
    })
}

/// Take the tape out of the thread-local context, consuming it.
///
/// Returns `None` if no tape exists (no tracked ops were recorded).
/// After this call, the next tracked op will create a fresh tape.
pub fn take_tape() -> Option<Tape> {
    CONTEXT.with(|c| c.borrow_mut().tape.take())
}

/// Install a pre-built tape into the thread-local context.
///
/// Used by the pipeline executor to set up per-micro-batch isolated tapes.
/// If a tape already exists, it is replaced (the old tape is dropped).
pub fn install_tape(tape: Tape) {
    CONTEXT.with(|c| {
        c.borrow_mut().tape = Some(tape);
    });
}

/// Returns `true` if tape recording is currently suppressed by a
/// [`NoGradGuard`].
pub(crate) fn is_no_grad() -> bool {
    CONTEXT.with(|c| c.borrow().no_grad_depth > 0)
}

// ---------------------------------------------------------------------------
// NoGradGuard — RAII suppression of tape recording
// ---------------------------------------------------------------------------

/// RAII guard that suppresses tape recording for the duration of its
/// lifetime.
///
/// Supports nesting: creating a second `NoGradGuard` while one is already
/// active increments the depth counter.  Recording resumes only when all
/// guards have been dropped (depth returns to 0).
///
/// # Example
///
/// ```ignore
/// let _guard = no_grad();
/// let y = x.add(&x);  // not recorded on the tape
/// // _guard drops here → recording resumes
/// ```
pub struct NoGradGuard;

/// Create a [`NoGradGuard`] that suppresses autograd tape recording.
///
/// The guard is active until it is dropped.  Nesting is supported.
pub fn no_grad() -> NoGradGuard {
    CONTEXT.with(|c| c.borrow_mut().no_grad_depth += 1);
    NoGradGuard
}

impl Drop for NoGradGuard {
    fn drop(&mut self) {
        CONTEXT.with(|c| {
            let mut ctx = c.borrow_mut();
            ctx.no_grad_depth = ctx.no_grad_depth.saturating_sub(1);
        });
    }
}