apr-cli 0.32.0

CLI tool for APR model inspection, debugging, and operations
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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158


// ════════════════════════════════════════════════════════════════════
// Coverage tests for print_cross_attention_flow, print_self_attention_flow,
// print_ffn_flow, and compute_stats (PMAT coverage gap)
// ════════════════════════════════════════════════════════════════════

#[cfg(test)]
mod flow_print_cross_tests {
    use super::*;

    // ── compute_stats ──────────────────────────────────────────────────

    #[test]
    fn test_compute_stats_empty() {
        let (min, max, mean, std) = compute_stats(&[]);
        assert_eq!(min, 0.0);
        assert_eq!(max, 0.0);
        assert_eq!(mean, 0.0);
        assert_eq!(std, 0.0);
    }

    #[test]
    fn test_compute_stats_single_element() {
        let (min, max, mean, std) = compute_stats(&[42.0]);
        assert_eq!(min, 42.0);
        assert_eq!(max, 42.0);
        assert_eq!(mean, 42.0);
        assert_eq!(std, 0.0);
    }

    #[test]
    fn test_compute_stats_uniform() {
        let data = vec![5.0, 5.0, 5.0, 5.0];
        let (min, max, mean, std) = compute_stats(&data);
        assert_eq!(min, 5.0);
        assert_eq!(max, 5.0);
        assert_eq!(mean, 5.0);
        assert_eq!(std, 0.0);
    }

    #[test]
    fn test_compute_stats_known_values() {
        // [1, 2, 3, 4, 5]: mean = 3.0, var = 2.0, std = sqrt(2.0)
        let data = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let (min, max, mean, std) = compute_stats(&data);
        assert_eq!(min, 1.0);
        assert_eq!(max, 5.0);
        assert!((mean - 3.0).abs() < 1e-5);
        assert!((std - 2.0_f32.sqrt()).abs() < 1e-4);
    }

    #[test]
    fn test_compute_stats_negative_values() {
        let data = vec![-3.0, -1.0, 0.0, 1.0, 3.0];
        let (min, max, mean, std) = compute_stats(&data);
        assert_eq!(min, -3.0);
        assert_eq!(max, 3.0);
        assert!((mean - 0.0).abs() < 1e-5);
        // var = (9+1+0+1+9)/5 = 4.0, std = 2.0
        assert!((std - 2.0).abs() < 1e-4);
    }

    #[test]
    fn test_compute_stats_large_range() {
        let data = vec![-1000.0, 0.0, 1000.0];
        let (min, max, mean, _std) = compute_stats(&data);
        assert_eq!(min, -1000.0);
        assert_eq!(max, 1000.0);
        assert!((mean - 0.0).abs() < 1e-3);
    }

    // ── print_cross_attention_flow ─────────────────────────────────────

    #[test]
    fn test_print_cross_attention_flow_no_cross_attn_layers() {
        // No cross-attention layers in the tensor names
        let tensor_names = vec![
            "model.layers.0.self_attn.q_proj.weight".to_string(),
            "model.layers.0.self_attn.k_proj.weight".to_string(),
        ];
        // Should print "No cross-attention layers found" without panicking
        print_cross_attention_flow(None, &tensor_names, None, false);
    }

    #[test]
    fn test_print_cross_attention_flow_empty_names() {
        print_cross_attention_flow(None, &[], None, false);
    }

    #[test]
    fn test_print_cross_attention_flow_with_cross_attn() {
        let tensor_names = vec![
            "decoder.layers.0.encoder_attn.q_proj.weight".to_string(),
            "decoder.layers.0.encoder_attn.k_proj.weight".to_string(),
            "decoder.layers.0.encoder_attn.v_proj.weight".to_string(),
            "decoder.layers.0.encoder_attn.out_proj.weight".to_string(),
            "decoder.layers.1.encoder_attn.q_proj.weight".to_string(),
        ];
        // Should render cross-attention diagram for 2 layers without panicking
        print_cross_attention_flow(None, &tensor_names, None, false);
    }

    #[test]
    fn test_print_cross_attention_flow_with_layer_filter() {
        let tensor_names = vec![
            "decoder.layers.0.encoder_attn.q_proj.weight".to_string(),
            "decoder.layers.1.encoder_attn.q_proj.weight".to_string(),
            "decoder.layers.2.encoder_attn.q_proj.weight".to_string(),
        ];
        // Filter to layer 1 only
        print_cross_attention_flow(None, &tensor_names, Some("layers.1"), false);
    }

    #[test]
    fn test_print_cross_attention_flow_verbose_no_reader() {
        let tensor_names = vec![
            "decoder.layers.0.encoder_attn.q_proj.weight".to_string(),
        ];
        // Verbose with no reader — should skip weight stats gracefully
        print_cross_attention_flow(None, &tensor_names, None, true);
    }

    #[test]
    fn test_print_cross_attention_flow_cross_attn_variant() {
        // Test with "cross_attn" naming convention (alternative to "encoder_attn")
        let tensor_names = vec![
            "model.layers.0.cross_attn.q_proj.weight".to_string(),
            "model.layers.0.cross_attn.k_proj.weight".to_string(),
        ];
        print_cross_attention_flow(None, &tensor_names, None, false);
    }

    // ── print_self_attention_flow ──────────────────────────────────────

    #[test]
    fn test_print_self_attention_flow_no_panic() {
        print_self_attention_flow(None, &[], None, false);
    }

    #[test]
    fn test_print_self_attention_flow_with_names() {
        let names = vec!["model.layers.0.self_attn.q_proj.weight".to_string()];
        print_self_attention_flow(None, &names, None, true);
    }

    // ── print_ffn_flow ─────────────────────────────────────────────────

    #[test]
    fn test_print_ffn_flow_no_panic() {
        print_ffn_flow(None, &[], None, false);
    }

    #[test]
    fn test_print_ffn_flow_with_names() {
        let names = vec!["model.layers.0.mlp.fc1.weight".to_string()];
        print_ffn_flow(None, &names, None, true);
    }
}