axonml 0.6.0

A complete ML/AI framework in pure Rust - PyTorch-equivalent functionality
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
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276

//! VULCAN — Mechanical Systems Diagnostic Model (~1.1M params)
//!
//! # File
//! `crates/axonml/src/hvac/vulcan.rs`
//!
//! # Author
//! Andrew Jewell Sr - AutomataNexus
//!
//! # Updated
//! March 8, 2026
//!
//! # Disclaimer
//! Use at own risk. This software is provided "as is", without warranty of any
//! kind, express or implied. The author and AutomataNexus shall not be held
//! liable for any damages arising from the use of this software.

use std::collections::HashMap;

use axonml_autograd::Variable;
use axonml_nn::{BatchNorm1d, Dropout, FFT1d, Linear, Module, Parameter, ReLU, Sequential};

use super::aquilo::concat_variables;

// =============================================================================
// Vulcan Model
// =============================================================================

/// Mechanical systems diagnostic model.
///
/// Architecture (Wide & Deep with FFT vibration analysis):
/// - Wide branch:   Linear(672, 256)
/// - Deep branch:   5-layer MLP with BatchNorm and Dropout
/// - Vibration:     FFT1d(672) → 337 freq bins → 2-layer MLP
/// - Fusion:        Concat(256+128+64=448) → Linear(448, 256) → ReLU
///
/// Input: (batch, 672) — flattened 96 timesteps × 7 features
/// Outputs: mech_fault(15), bearing_health(4), vib_severity(4), rul(1)
pub struct Vulcan {
    wide_branch: Sequential,
    deep_branch: Sequential,
    fft: FFT1d,
    vib_linear1: Linear,
    vib_linear2: Linear,
    fusion: Sequential,
    mech_fault_head: Linear,
    bearing_health_head: Linear,
    vib_severity_head: Linear,
    rul_head: Linear,
    training: bool,
}

impl Default for Vulcan {
    fn default() -> Self {
        Self::new()
    }
}

impl Vulcan {
    /// Creates a new Vulcan model.
    pub fn new() -> Self {
        // Wide branch: Linear(672, 256)
        let wide_branch = Sequential::new().add(Linear::new(672, 256));

        // Deep branch: 5 layers with BN/ReLU/Dropout
        let deep_branch = Sequential::new()
            .add(Linear::new(672, 512))
            .add(BatchNorm1d::new(512))
            .add(ReLU)
            .add(Dropout::new(0.2))
            .add(Linear::new(512, 512))
            .add(BatchNorm1d::new(512))
            .add(ReLU)
            .add(Dropout::new(0.2))
            .add(Linear::new(512, 256))
            .add(BatchNorm1d::new(256))
            .add(ReLU)
            .add(Dropout::new(0.2))
            .add(Linear::new(256, 256))
            .add(BatchNorm1d::new(256))
            .add(ReLU)
            .add(Linear::new(256, 128))
            .add(ReLU);

        // FFT on full 672-length input → 337 frequency bins
        let fft = FFT1d::new(672);

        // Vibration branch linear layers (applied after FFT)
        let vib_linear1 = Linear::new(337, 256);
        let vib_linear2 = Linear::new(256, 64);

        // Fusion: concat wide(256) + deep(128) + vibration(64) = 448
        let fusion = Sequential::new().add(Linear::new(448, 256)).add(ReLU);

        // Output heads
        let mech_fault_head = Linear::new(256, 15);
        let bearing_health_head = Linear::new(256, 4);
        let vib_severity_head = Linear::new(256, 4);
        let rul_head = Linear::new(256, 1);

        Self {
            wide_branch,
            deep_branch,
            fft,
            vib_linear1,
            vib_linear2,
            fusion,
            mech_fault_head,
            bearing_health_head,
            vib_severity_head,
            rul_head,
            training: true,
        }
    }

    /// Forward pass returning all output heads.
    ///
    /// Returns (mech_fault, bearing_health, vib_severity, rul, embedding)
    pub fn forward_all(
        &self,
        input: &Variable,
    ) -> (Variable, Variable, Variable, Variable, Variable) {
        let shape = input.shape();
        let batch = shape[0];

        // Wide branch: (batch, 672) → (batch, 256)
        let wide_out = self.wide_branch.forward(input);

        // Deep branch: (batch, 672) → (batch, 128)
        let deep_out = self.deep_branch.forward(input);

        // Vibration branch: FFT → Linear → ReLU → Linear
        let fft_out = self.fft.forward(input); // (batch, 337)
        let vib_hidden = self.vib_linear1.forward(&fft_out); // (batch, 256)
        let vib_hidden = vib_hidden.relu(); // ReLU
        let vib_out = self.vib_linear2.forward(&vib_hidden); // (batch, 64)

        // Fusion: concat wide(256) + deep(128) + vibration(64) = 448
        let fused = concat_variables(&[&wide_out, &deep_out, &vib_out], batch);
        let embedding = self.fusion.forward(&fused); // (batch, 256)

        // Output heads
        let mech_fault = self.mech_fault_head.forward(&embedding);
        let bearing_health = self.bearing_health_head.forward(&embedding);
        let vib_severity = self.vib_severity_head.forward(&embedding);
        let rul = self.rul_head.forward(&embedding);

        (mech_fault, bearing_health, vib_severity, rul, embedding)
    }

    /// Returns the embedding dimension for downstream aggregators.
    pub fn embedding_dim() -> usize {
        256
    }

    /// Returns total output dimension (15 + 4 + 4 + 1 = 24).
    pub fn output_dim() -> usize {
        24
    }
}

impl Module for Vulcan {
    fn forward(&self, input: &Variable) -> Variable {
        let (mech_fault, _, _, _, _) = self.forward_all(input);
        mech_fault
    }

    fn parameters(&self) -> Vec<Parameter> {
        let mut params = Vec::new();
        params.extend(self.wide_branch.parameters());
        params.extend(self.deep_branch.parameters());
        params.extend(self.vib_linear1.parameters());
        params.extend(self.vib_linear2.parameters());
        params.extend(self.fusion.parameters());
        params.extend(self.mech_fault_head.parameters());
        params.extend(self.bearing_health_head.parameters());
        params.extend(self.vib_severity_head.parameters());
        params.extend(self.rul_head.parameters());
        params
    }

    fn named_parameters(&self) -> HashMap<String, Parameter> {
        let mut params = HashMap::new();
        for (n, p) in self.wide_branch.named_parameters() {
            params.insert(format!("wide_branch.{n}"), p);
        }
        for (n, p) in self.deep_branch.named_parameters() {
            params.insert(format!("deep_branch.{n}"), p);
        }
        for (n, p) in self.vib_linear1.named_parameters() {
            params.insert(format!("vib_linear1.{n}"), p);
        }
        for (n, p) in self.vib_linear2.named_parameters() {
            params.insert(format!("vib_linear2.{n}"), p);
        }
        for (n, p) in self.fusion.named_parameters() {
            params.insert(format!("fusion.{n}"), p);
        }
        for (n, p) in self.mech_fault_head.named_parameters() {
            params.insert(format!("mech_fault_head.{n}"), p);
        }
        for (n, p) in self.bearing_health_head.named_parameters() {
            params.insert(format!("bearing_health_head.{n}"), p);
        }
        for (n, p) in self.vib_severity_head.named_parameters() {
            params.insert(format!("vib_severity_head.{n}"), p);
        }
        for (n, p) in self.rul_head.named_parameters() {
            params.insert(format!("rul_head.{n}"), p);
        }
        params
    }

    fn set_training(&mut self, training: bool) {
        self.training = training;
        self.wide_branch.set_training(training);
        self.deep_branch.set_training(training);
        self.fusion.set_training(training);
    }

    fn is_training(&self) -> bool {
        self.training
    }

    fn name(&self) -> &'static str {
        "Vulcan"
    }
}

// =============================================================================
// Tests
// =============================================================================

#[cfg(test)]
mod tests {
    use super::*;
    use axonml_tensor::Tensor;

    #[test]
    fn test_vulcan_output_shapes() {
        let model = Vulcan::new();
        let input = Variable::new(
            Tensor::from_vec(vec![1.0; 2 * 672], &[2, 672]).unwrap(),
            false,
        );
        let (mech_fault, bearing_health, vib_severity, rul, emb) = model.forward_all(&input);

        assert_eq!(mech_fault.shape(), vec![2, 15]);
        assert_eq!(bearing_health.shape(), vec![2, 4]);
        assert_eq!(vib_severity.shape(), vec![2, 4]);
        assert_eq!(rul.shape(), vec![2, 1]);
        assert_eq!(emb.shape(), vec![2, 256]);
    }

    #[test]
    fn test_vulcan_parameter_count() {
        let model = Vulcan::new();
        let total: usize = model.parameters().iter().map(|p| p.numel()).sum();
        // Expected ~1.1M params
        assert!(
            total > 900_000 && total < 1_300_000,
            "Vulcan has {} params, expected ~1.1M",
            total
        );
    }

    #[test]
    fn test_vulcan_forward_module_trait() {
        let model = Vulcan::new();
        let input = Variable::new(
            Tensor::from_vec(vec![1.0; 4 * 672], &[4, 672]).unwrap(),
            false,
        );
        let output = model.forward(&input);
        assert_eq!(output.shape(), vec![4, 15]);
    }
}