onnx-ir 0.19.0

ONNX-IR is a pure Rust library for parsing ONNX models into an intermediate representation that can be used to generate code for various ML/DL frameworks
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
133

use crate::from_onnx::element_type_from_proto;
use crate::ir::{ArgType, ElementType, Node, TensorType};

/// Configuration for EyeLike operations
#[derive(Debug, Clone, new)]
pub struct EyeLikeConfig {
    /// Data type of the output tensor (optional, defaults to input type)
    pub dtype: Option<ElementType>,
    /// Diagonal offset (0 = main diagonal, >0 = upper, <0 = lower)
    pub k: i64,
}

/// Create an EyeLike configuration from the node
pub fn eye_like_config(node: &Node) -> EyeLikeConfig {
    let mut dtype = None;
    let mut k = 0i64; // default to main diagonal

    // Extract attributes
    for (key, value) in node.attrs.iter() {
        match key.as_str() {
            "dtype" => {
                let dtype_i32 = value.clone().into_i32();
                dtype = Some(
                    element_type_from_proto(dtype_i32)
                        .unwrap_or_else(|e| panic!("Unsupported dtype for EyeLike: {}", e)),
                );
            }
            "k" => {
                k = value.clone().into_i64();
            }
            _ => {}
        }
    }

    EyeLikeConfig { dtype, k }
}

/// Update output for EyeLike - output has same shape as input, but may have different dtype
pub fn eye_like_update_output(node: &mut Node) {
    log::debug!("EyeLike rank inference for node {}", node.name);

    match &node.inputs[0].ty {
        ArgType::Tensor(tensor) => {
            assert_eq!(tensor.rank, 2, "Input rank must be 2D tensor");

            let config = eye_like_config(node);
            // Output type is either specified dtype or input type
            let output_type = config.dtype.unwrap_or_else(|| tensor.elem_type.clone());

            node.outputs[0].ty = ArgType::Tensor(TensorType {
                elem_type: output_type,
                rank: tensor.rank,
                static_shape: tensor.static_shape.clone(),
            });
            log::debug!("EyeLike output tensor rank: {}", tensor.rank);
        }
        _ => panic!("EyeLike operation requires 2D tensor input"),
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::ir::{ElementType, NodeType};
    use crate::node::test_utils::NodeBuilder;
    use crate::protos::tensor_proto::DataType;
    use protobuf::Enum;

    #[test]
    fn test_eye_like_update_output() {
        let mut node = NodeBuilder::new(NodeType::EyeLike, "test_eye_like")
            .input_tensor_f32("input", 2, Some(vec![3, 3]))
            .output_tensor_f32("output", 2, None) // rank will be updated
            .build();

        eye_like_update_output(&mut node);

        match &node.outputs[0].ty {
            ArgType::Tensor(tensor) => {
                assert_eq!(tensor.elem_type, ElementType::Float32);
                assert_eq!(tensor.rank, 2);
                assert_eq!(tensor.static_shape, Some(vec![3, 3]));
            }
            _ => panic!("Expected tensor output"),
        }
    }

    #[test]
    fn test_eye_like_config_default() {
        let node = NodeBuilder::new(NodeType::EyeLike, "test_eye_like")
            .input_tensor_f32("input", 2, Some(vec![4, 4]))
            .output_tensor_f32("output", 2, None)
            .build();

        let config = eye_like_config(&node);
        assert_eq!(config.k, 0);
        assert_eq!(config.dtype, None);
    }

    #[test]
    fn test_eye_like_config_with_attributes() {
        let node = NodeBuilder::new(NodeType::EyeLike, "test_eye_like")
            .input_tensor_f32("input", 2, Some(vec![4, 4]))
            .output_tensor_f32("output", 2, None)
            .attr_int("k", -1)
            .attr_int("dtype", DataType::INT64.value() as i64)
            .build();

        let config = eye_like_config(&node);
        assert_eq!(config.k, -1);
        assert_eq!(config.dtype, Some(ElementType::Int64));
    }

    #[test]
    fn test_eye_like_update_output_with_dtype() {
        let mut node = NodeBuilder::new(NodeType::EyeLike, "test_eye_like")
            .input_tensor_f32("input", 2, Some(vec![3, 3]))
            .output_tensor_f32("output", 2, None)
            .attr_int("dtype", DataType::INT32.value() as i64)
            .build();

        eye_like_update_output(&mut node);

        match &node.outputs[0].ty {
            ArgType::Tensor(tensor) => {
                assert_eq!(tensor.elem_type, ElementType::Int32);
                assert_eq!(tensor.rank, 2);
                assert_eq!(tensor.static_shape, Some(vec![3, 3]));
            }
            _ => panic!("Expected tensor output"),
        }
    }
}