cortiq-core 0.5.78

CMF (Cortiq Model Format): a self-describing, memory-mappable binary container for quantized LLM weights, tokenizer, per-task masks and per-skill delta records.
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

//! Standalone skill files at the container level: the identity keys
//! round-trip, the feature bit raises itself from the record, and a
//! reader treats the file as data — not as a network missing most of
//! its organs.

use cortiq_core::format::features;
use cortiq_core::types::{LayerType, ModelArch, NormStyle};
use cortiq_core::{CmfHeader, CmfModel, SkillRecord, TensorDtype, TensorSpec};

fn tiny_header() -> CmfHeader {
    // The loop_masks fixture's arch, wrapped in the smallest header the
    // writer accepts.
    let arch = ModelArch {
        arch_name: "tiny-looped".into(),
        hidden_size: 8,
        intermediate_size: 16,
        num_layers: 2,
        num_attention_heads: 2,
        num_kv_heads: 1,
        head_dim: 4,
        vocab_size: 10,
        layer_types: vec![LayerType::FullAttention; 2],
        rms_norm_eps: 1e-6,
        norm_style: NormStyle::Qwen,
        rope_theta: 10_000.0,
        tie_word_embeddings: false,
        partial_rotary_factor: 1.0,
        yarn: None,
        attention_heads_per_layer: None,
        local_partial_rotary_factor: None,
        mtp: None,
        moe: None,
        linear_core: None,
        max_position_embeddings: 64,
        linear_conv_kernel_dim: None,
        linear_num_key_heads: None,
        linear_num_value_heads: None,
        linear_key_head_dim: None,
        linear_value_head_dim: None,
        hidden_act: "silu".into(),
        embed_multiplier: 1.0,
        query_pre_attn_scalar: None,
        sliding_window: None,
        sliding_window_pattern: None,
        rope_local_base_freq: None,
        global_head_dim: None,
        num_global_kv_heads: None,
        global_partial_rotary_factor: None,
        final_logit_softcapping: None,
        attn_logit_softcapping: None,
        mla: None,
        activation_situ_beta: None,
        activation_situ_linear_beta: None,
        attn_v_norm: false,
        num_loops: 1,
        kda_gate_lower_bound: None,
        g3n: None,
        rope_freq_factors: None,
        logit_multiplier: None,
        loop_final_norm: false,
    };
    let mut h: CmfHeader = serde_json::from_value(serde_json::json!({
        "version": 2,
        "arch": serde_json::to_value(&arch).unwrap(),
        "quant_type": "F16",
    }))
    .expect("tiny header");
    h.arch = arch;
    h
}

fn t(name: &str, fill: u8) -> TensorSpec {
    TensorSpec {
        name: name.into(),
        dtype: TensorDtype::F16,
        shape: vec![4, 4],
        data: vec![fill; 32],
    }
}

#[test]
fn skill_keys_round_trip_and_raise_the_feature_bit() {
    let dir = std::env::temp_dir().join(format!("cmf-skill-test-{}", std::process::id()));
    std::fs::create_dir_all(&dir).unwrap();
    let base_p = dir.join("base.cmf");
    let skill_p = dir.join("gfx.skill.cmf");

    // A base with three tensors.
    let header = tiny_header();
    CmfModel::write(
        &base_p,
        &header,
        &[t("model.layers.0.mlp.down_proj.weight", 1), t("model.layers.1.mlp.down_proj.weight", 2), t("lm_head.weight", 3)],
        None,
        None,
    )
    .unwrap();
    let base = CmfModel::open(&base_p).unwrap();
    assert_eq!(
        base.required_features & features::SKILL_FILE,
        0,
        "a plain model must not carry the skill bit"
    );

    // A skill file: one replaced tensor + the identity keys.
    let mut sh = tiny_header();
    sh.skills = vec![SkillRecord {
        id: "gfx".into(),
        name: Some("graphics".into()),
        layers: vec![1],
        selection: None,
        input_mask_task: None,
        quality: None,
        base_dir_hash: Some(format!("{:016x}", base.dir_hash())),
        base_arch: Some("tiny-looped".into()),
        task: Some("specialist".into()),
        provenance: Some(serde_json::json!({"corpus": "test"})),
    }];
    CmfModel::write(
        &skill_p,
        &sh,
        &[t("model.layers.1.mlp.down_proj.weight", 9)],
        None,
        None,
    )
    .unwrap();

    let skill = CmfModel::open(&skill_p).unwrap();
    assert_ne!(
        skill.required_features & features::SKILL_FILE,
        0,
        "a bound skill record must raise the SKILL_FILE bit"
    );
    let rec = &skill.header.skills[0];
    assert_eq!(rec.base_dir_hash.as_deref(), Some(format!("{:016x}", base.dir_hash()).as_str()));
    assert_eq!(rec.base_arch.as_deref(), Some("tiny-looped"));
    assert_eq!(rec.task.as_deref(), Some("specialist"));
    assert_eq!(rec.layers, vec![1]);
    // The partial tensor set is exactly what was cut.
    assert_eq!(skill.tensors.len(), 1);
    assert_eq!(skill.tensors[0].name, "model.layers.1.mlp.down_proj.weight");

    std::fs::remove_dir_all(&dir).ok();
}