speakrs 0.3.2

Fast Rust speaker diarization with pyannote-level accuracy and native CoreML/CUDA acceleration
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

use std::path::{Path, PathBuf};

#[cfg(feature = "online")]
use crate::inference::ExecutionMode;

const SEGMENTATION_ONNX: &str = "segmentation-3.0.onnx";
const EMBEDDING_ONNX: &str = "wespeaker-voxceleb-resnet34.onnx";

/// Resolved model paths for the speakrs pipeline
///
/// Captures the three root paths needed by [`SegmentationModel`], [`EmbeddingModel`],
/// and `PldaTransform`. Variant models (batched, CoreML, split) are derived
/// internally by each model constructor from the base ONNX path.
///
/// [`SegmentationModel`]: crate::inference::segmentation::SegmentationModel
/// [`EmbeddingModel`]: crate::inference::embedding::EmbeddingModel
#[derive(Debug, Clone)]
pub struct ModelBundle {
    segmentation_onnx: PathBuf,
    embedding_onnx: PathBuf,
    plda_dir: PathBuf,
}

impl ModelBundle {
    /// Resolve paths from a local directory containing all model files
    pub fn from_dir(models_dir: impl Into<PathBuf>) -> Self {
        let dir = models_dir.into();
        Self {
            segmentation_onnx: dir.join(SEGMENTATION_ONNX),
            embedding_onnx: dir.join(EMBEDDING_ONNX),
            plda_dir: dir,
        }
    }

    /// Download models from HuggingFace and resolve paths
    #[cfg(feature = "online")]
    pub fn from_pretrained(mode: ExecutionMode) -> Result<Self, hf_hub::api::sync::ApiError> {
        let manager = ModelManager::new()?;
        let dir = manager.ensure(mode)?;
        Ok(Self::from_dir(dir))
    }

    /// Base ONNX path for the segmentation model
    pub fn segmentation_path(&self) -> &Path {
        &self.segmentation_onnx
    }

    /// Base ONNX path for the embedding model
    pub fn embedding_path(&self) -> &Path {
        &self.embedding_onnx
    }

    /// Directory containing PLDA parameter files
    pub fn plda_dir(&self) -> &Path {
        &self.plda_dir
    }
}

#[cfg(feature = "online")]
const HF_REPO: &str = "avencera/speakrs-models";

/// Manages downloading and caching speakrs ONNX models from HuggingFace
#[cfg(feature = "online")]
pub struct ModelManager {
    repo: hf_hub::api::sync::ApiRepo,
}

#[cfg(feature = "online")]
impl ModelManager {
    /// Create a manager using the default HuggingFace cache directory
    pub fn new() -> Result<Self, hf_hub::api::sync::ApiError> {
        let api = hf_hub::api::sync::Api::new()?;
        let repo = api.model(HF_REPO.to_string());
        Ok(Self { repo })
    }

    /// Create a manager with a custom cache directory
    pub fn with_cache_dir(cache_dir: PathBuf) -> Result<Self, hf_hub::api::sync::ApiError> {
        let api =
            hf_hub::api::sync::ApiBuilder::from_cache(hf_hub::Cache::new(cache_dir)).build()?;
        let repo = api.model(HF_REPO.to_string());
        Ok(Self { repo })
    }

    /// Download a single file, returns path to cached copy
    pub fn get(&self, filename: impl AsRef<str>) -> Result<PathBuf, hf_hub::api::sync::ApiError> {
        self.repo.get(filename.as_ref())
    }

    /// Ensure all files for a mode are downloaded, return base models dir
    pub fn ensure(&self, mode: ExecutionMode) -> Result<PathBuf, hf_hub::api::sync::ApiError> {
        let files = required_files(mode);
        for file in &files {
            self.repo.get(file)?;
        }
        // all files land in the same snapshot dir
        let first = self.repo.get(&files[0])?;
        let Some(parent) = first.parent() else {
            return Ok(first);
        };
        Ok(parent.to_path_buf())
    }
}

#[cfg(feature = "online")]
const PLDA_FILES: &[&str] = &[
    "plda_lda.npy",
    "plda_tr.npy",
    "plda_mu.npy",
    "plda_psi.npy",
    "plda_mean1.npy",
    "plda_mean2.npy",
    "wespeaker-voxceleb-resnet34.min_num_samples.txt",
];

#[cfg(feature = "online")]
const ONNX_FILES: &[&str] = &[
    "segmentation-3.0.onnx",
    "wespeaker-voxceleb-resnet34.onnx",
    "wespeaker-voxceleb-resnet34.onnx.data",
];

#[cfg(feature = "online")]
fn mlmodelc_files(name: &str) -> Vec<String> {
    vec![
        format!("{name}/model.mil"),
        format!("{name}/coremldata.bin"),
        format!("{name}/weights/weight.bin"),
        format!("{name}/analytics/coremldata.bin"),
    ]
}

#[cfg(feature = "online")]
fn extend_mlmodelc_files(files: &mut Vec<String>, names: &[&str]) {
    for name in names {
        files.extend(mlmodelc_files(name));
    }
}

#[cfg(feature = "online")]
const COREML_COMMON_MODEL_STEMS: &[&str] = &[
    "segmentation-3.0.mlmodelc",
    "segmentation-3.0-b32.mlmodelc",
    "segmentation-3.0-b64.mlmodelc",
    "wespeaker-fbank.mlmodelc",
    "wespeaker-fbank-b32.mlmodelc",
    "wespeaker-fbank-30s.mlmodelc",
    "wespeaker-multimask-tail-b32.mlmodelc",
    "wespeaker-voxceleb-resnet34-tail.mlmodelc",
    "wespeaker-voxceleb-resnet34-tail-b3.mlmodelc",
    "wespeaker-voxceleb-resnet34-tail-b32.mlmodelc",
];

#[cfg(feature = "online")]
const COREML_CHUNK_MODEL_STEMS: &[&str] = &[
    "wespeaker-chunk-emb-s12-w22.mlmodelc",
    "wespeaker-chunk-emb-s12-w37.mlmodelc",
    "wespeaker-chunk-emb-s12-w53.mlmodelc",
    "wespeaker-chunk-emb-s12-w84.mlmodelc",
    "wespeaker-chunk-emb-s12-w116.mlmodelc",
];

#[cfg(feature = "online")]
const COREML_FAST_SEGMENTATION_MODEL_STEMS: &[&str] = &[
    "segmentation-3.0-w8a16.mlmodelc",
    "segmentation-3.0-b32-w8a16.mlmodelc",
    "segmentation-3.0-b64-w8a16.mlmodelc",
];

#[cfg(feature = "online")]
const COREML_FAST_CHUNK_MODEL_STEMS: &[&str] = &[
    "wespeaker-chunk-emb-s25-w11.mlmodelc",
    "wespeaker-chunk-emb-s25-w16.mlmodelc",
    "wespeaker-chunk-emb-s25-w21.mlmodelc",
    "wespeaker-chunk-emb-s25-w26.mlmodelc",
    "wespeaker-chunk-emb-s25-w36.mlmodelc",
    "wespeaker-chunk-emb-s25-w46.mlmodelc",
    "wespeaker-chunk-emb-s25-w56.mlmodelc",
];

#[cfg(feature = "online")]
fn required_files(mode: ExecutionMode) -> Vec<String> {
    let mut files: Vec<String> = PLDA_FILES.iter().map(|s| s.to_string()).collect();

    match mode {
        ExecutionMode::Cpu => {
            files.extend(ONNX_FILES.iter().map(|s| s.to_string()));
        }
        ExecutionMode::Cuda | ExecutionMode::CudaFast => {
            files.extend(ONNX_FILES.iter().map(|s| s.to_string()));
            // split models for multi-mask embedding (CPU fbank + GPU multi-mask)
            files.push("wespeaker-fbank.onnx".to_string());
            files.push("wespeaker-fbank-b32.onnx".to_string());
            files.push("wespeaker-multimask-tail.onnx".to_string());
            files.push("wespeaker-multimask-tail-b32.onnx".to_string());
            // batched seg/emb models
            files.push("segmentation-3.0-b32.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-b64.onnx".to_string());
        }
        ExecutionMode::CoreMl => {
            // CoreML modes still need the ONNX segmentation model for the constructor
            files.push("segmentation-3.0.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34.onnx.data".to_string());
            // b32 batched ONNX for segmentation
            files.push("segmentation-3.0-b32.onnx".to_string());
            // split ONNX models for embedding
            files.push("wespeaker-fbank.onnx".to_string());
            files.push("wespeaker-fbank-b32.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-tail.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-tail-b3.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-tail-b32.onnx".to_string());
            extend_mlmodelc_files(&mut files, COREML_COMMON_MODEL_STEMS);
            extend_mlmodelc_files(&mut files, COREML_CHUNK_MODEL_STEMS);
        }
        ExecutionMode::CoreMlFast => {
            files.push("segmentation-3.0.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34.onnx.data".to_string());
            files.push("segmentation-3.0-b32.onnx".to_string());
            files.push("wespeaker-fbank.onnx".to_string());
            files.push("wespeaker-fbank-b32.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-tail.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-tail-b3.onnx".to_string());
            files.push("wespeaker-voxceleb-resnet34-tail-b32.onnx".to_string());
            extend_mlmodelc_files(&mut files, COREML_COMMON_MODEL_STEMS);
            extend_mlmodelc_files(&mut files, COREML_FAST_SEGMENTATION_MODEL_STEMS);
            extend_mlmodelc_files(&mut files, COREML_FAST_CHUNK_MODEL_STEMS);
        }
    }

    files
}

#[cfg(all(test, feature = "online"))]
mod tests {
    use super::*;

    #[test]
    fn coreml_required_files_include_chunk_fast_path_assets() {
        let files = required_files(ExecutionMode::CoreMl);
        assert!(files.contains(&"segmentation-3.0-b64.mlmodelc/model.mil".to_string()));
        assert!(files.contains(&"wespeaker-fbank-30s.mlmodelc/model.mil".to_string()));
        assert!(files.contains(&"wespeaker-multimask-tail-b32.mlmodelc/model.mil".to_string()));
        assert!(files.contains(&"wespeaker-chunk-emb-s12-w116.mlmodelc/model.mil".to_string()));
    }

    #[test]
    fn coreml_fast_required_files_include_fast_assets() {
        let files = required_files(ExecutionMode::CoreMlFast);
        assert!(files.contains(&"segmentation-3.0-w8a16.mlmodelc/model.mil".to_string()));
        assert!(files.contains(&"segmentation-3.0-b64-w8a16.mlmodelc/model.mil".to_string()));
        assert!(files.contains(&"wespeaker-chunk-emb-s25-w56.mlmodelc/model.mil".to_string()));
    }
}