onnx-runtime-shape-inference 0.1.0-dev.2

Symbolic shape inference for the ORT 2.0 runtime: an extensible, opset-aware per-op registry with symbolic dimension arithmetic and shape-data propagation over onnx-runtime-ir graphs
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

//! The result summary returned by whole-graph inference.

use onnx_runtime_ir::ValueId;

/// A summary of what whole-graph inference resolved.
///
/// A value is "resolved" once it has a known dtype and a known-rank shape (every
/// dimension is concrete or symbolic — never unknown). Values an op rule could
/// not resolve (unregistered op, data-dependent extent without shape-data) are
/// listed in [`unresolved`](InferenceReport::unresolved).
#[derive(Clone, Debug, Default, PartialEq, Eq)]
pub struct InferenceReport {
    /// Total number of live values in the graph.
    pub total_values: usize,
    /// Values that ended with a resolved shape.
    pub resolved: Vec<ValueId>,
    /// Values left without a resolved shape.
    pub unresolved: Vec<ValueId>,
    /// Number of fresh symbolic dimensions minted during inference.
    pub fresh_symbols: usize,
}

impl InferenceReport {
    /// The number of resolved values.
    pub fn num_resolved(&self) -> usize {
        self.resolved.len()
    }

    /// The number of unresolved values.
    pub fn num_unresolved(&self) -> usize {
        self.unresolved.len()
    }

    /// Whether every live value in the graph resolved to a known shape.
    pub fn fully_resolved(&self) -> bool {
        self.unresolved.is_empty()
    }
}