|
| 1 | +use std::collections::HashSet; |
| 2 | +use std::path::Path; |
| 3 | +use std::time::{Duration, SystemTime}; |
| 4 | + |
| 5 | +use cpu_time::ProcessTime; |
| 6 | +use diskann::index_view::IndexView; |
| 7 | +use system::vector_data::VectorData; |
| 8 | +use system::vector_point::VectorPoint; |
| 9 | +use tokio::fs::File; |
| 10 | +use tokio::io::{self, AsyncReadExt, BufReader}; |
| 11 | + |
| 12 | +#[derive(Debug)] |
| 13 | +pub struct SiftDataset { |
| 14 | + pub dimension: u32, |
| 15 | + pub vectors: Vec<VectorData>, |
| 16 | +} |
| 17 | + |
| 18 | +impl SiftDataset { |
| 19 | + pub async fn from_fvecs<P: AsRef<Path>>(path: P) -> io::Result<Self> { |
| 20 | + let file = File::open(path).await?; |
| 21 | + let mut reader = BufReader::new(file); |
| 22 | + |
| 23 | + let mut vectors = Vec::new(); |
| 24 | + let mut dimension = 0u32; |
| 25 | + let mut buffer = [0u8; 4]; |
| 26 | + |
| 27 | + if reader.read_exact(&mut buffer).await.is_ok() { |
| 28 | + dimension = u32::from_le_bytes(buffer); |
| 29 | + |
| 30 | + // Read first vector |
| 31 | + let mut vec = vec![0f32; dimension as usize]; |
| 32 | + for item in vec.iter_mut().take(dimension as usize) { |
| 33 | + reader.read_exact(&mut buffer).await?; |
| 34 | + *item = f32::from_le_bytes(buffer); |
| 35 | + } |
| 36 | + println!("vector: {:?}", vec); |
| 37 | + let vec_data = VectorData::from_f32(vec); |
| 38 | + vectors.push(vec_data); |
| 39 | + } |
| 40 | + |
| 41 | + loop { |
| 42 | + match reader.read_exact(&mut buffer).await { |
| 43 | + Ok(_) => { |
| 44 | + let dim = u32::from_le_bytes(buffer); |
| 45 | + if dim != dimension { |
| 46 | + return Err(io::Error::new( |
| 47 | + io::ErrorKind::InvalidData, |
| 48 | + "Inconsistent vector dimensions", |
| 49 | + )); |
| 50 | + } |
| 51 | + |
| 52 | + let mut vec = vec![0f32; dimension as usize]; |
| 53 | + for item in vec.iter_mut().take(dimension as usize) { |
| 54 | + reader.read_exact(&mut buffer).await?; |
| 55 | + *item = f32::from_le_bytes(buffer); |
| 56 | + } |
| 57 | + let vec_data = VectorData::from_f32(vec); |
| 58 | + vectors.push(vec_data); |
| 59 | + } |
| 60 | + Err(e) if e.kind() == io::ErrorKind::UnexpectedEof => break, |
| 61 | + Err(e) => return Err(e), |
| 62 | + } |
| 63 | + } |
| 64 | + |
| 65 | + Ok(SiftDataset { dimension, vectors }) |
| 66 | + } |
| 67 | + |
| 68 | + pub async fn from_bvecs<P: AsRef<Path>>(path: P) -> io::Result<Self> { |
| 69 | + let file = File::open(path).await?; |
| 70 | + let mut reader = BufReader::new(file); |
| 71 | + |
| 72 | + let mut vectors = Vec::new(); |
| 73 | + let mut dimension = 0u32; |
| 74 | + let mut buffer = [0u8; 4]; |
| 75 | + |
| 76 | + if reader.read_exact(&mut buffer).await.is_ok() { |
| 77 | + dimension = u32::from_le_bytes(buffer); |
| 78 | + |
| 79 | + let mut vec = vec![0u8; dimension as usize]; |
| 80 | + reader.read_exact(&mut vec).await?; |
| 81 | + let vec = vec.iter().map(|&x| x as f32).collect::<Vec<f32>>(); |
| 82 | + let vec_data = VectorData::from_f32(vec); |
| 83 | + vectors.push(vec_data); |
| 84 | + } |
| 85 | + |
| 86 | + loop { |
| 87 | + match reader.read_exact(&mut buffer).await { |
| 88 | + Ok(_) => { |
| 89 | + let dim = u32::from_le_bytes(buffer); |
| 90 | + if dim != dimension { |
| 91 | + return Err(io::Error::new( |
| 92 | + io::ErrorKind::InvalidData, |
| 93 | + "Inconsistent vector dimensions", |
| 94 | + )); |
| 95 | + } |
| 96 | + |
| 97 | + let mut vec = vec![0u8; dimension as usize]; |
| 98 | + reader.read_exact(&mut vec).await?; |
| 99 | + let vec = vec.iter().map(|&x| x as f32).collect::<Vec<f32>>(); |
| 100 | + let vec_data = VectorData::from_f32(vec); |
| 101 | + vectors.push(vec_data); |
| 102 | + } |
| 103 | + Err(e) if e.kind() == io::ErrorKind::UnexpectedEof => break, |
| 104 | + Err(e) => return Err(e), |
| 105 | + } |
| 106 | + } |
| 107 | + |
| 108 | + Ok(SiftDataset { dimension, vectors }) |
| 109 | + } |
| 110 | + |
| 111 | + pub async fn from_ivecs<P: AsRef<Path>>(path: P) -> io::Result<Vec<Vec<u32>>> { |
| 112 | + let file = File::open(path).await?; |
| 113 | + let mut reader = BufReader::new(file); |
| 114 | + |
| 115 | + let mut results = Vec::new(); |
| 116 | + let mut buffer = [0u8; 4]; |
| 117 | + |
| 118 | + loop { |
| 119 | + match reader.read_exact(&mut buffer).await { |
| 120 | + Ok(_) => { |
| 121 | + let k = u32::from_le_bytes(buffer); |
| 122 | + let mut neighbors = vec![0u32; k as usize]; |
| 123 | + |
| 124 | + for item in neighbors.iter_mut().take(k as usize) { |
| 125 | + reader.read_exact(&mut buffer).await?; |
| 126 | + *item = u32::from_le_bytes(buffer); |
| 127 | + } |
| 128 | + results.push(neighbors); |
| 129 | + } |
| 130 | + Err(e) if e.kind() == io::ErrorKind::UnexpectedEof => break, |
| 131 | + Err(e) => return Err(e), |
| 132 | + } |
| 133 | + } |
| 134 | + |
| 135 | + Ok(results) |
| 136 | + } |
| 137 | + |
| 138 | + pub fn get_vector(&self, idx: usize) -> Option<&VectorData> { |
| 139 | + self.vectors.get(idx) |
| 140 | + } |
| 141 | +} |
| 142 | + |
| 143 | +pub fn compute_recall(results: &[Vec<u32>], ground_truth: &[Vec<u32>], k: usize) -> f64 { |
| 144 | + assert_eq!(results.len(), ground_truth.len(), "Results and groundtruth must have same length"); |
| 145 | + |
| 146 | + let num_queries = results.len(); |
| 147 | + let mut total_matches = 0; |
| 148 | + |
| 149 | + for (result, truth) in results.iter().zip(ground_truth.iter()) { |
| 150 | + let truth_set: HashSet<u32> = truth.iter().take(k).copied().collect(); |
| 151 | + |
| 152 | + let matches = result.iter().take(k).filter(|&&idx| truth_set.contains(&idx)).count(); |
| 153 | + println!("Matches: {}", matches); |
| 154 | + total_matches += matches; |
| 155 | + } |
| 156 | + |
| 157 | + total_matches as f64 / (num_queries * k) as f64 |
| 158 | +} |
| 159 | + |
| 160 | +async fn compute_recall_at_k( |
| 161 | + index_view: &IndexView, query: &SiftDataset, ground_truth: &[Vec<u32>], k: usize, |
| 162 | +) { |
| 163 | + let mut results = Vec::new(); |
| 164 | + for q in query.vectors.iter() { |
| 165 | + let res = index_view.search(q, k, 128).await; |
| 166 | + results.push(res); |
| 167 | + } |
| 168 | + |
| 169 | + let recall = compute_recall(&results, ground_truth, k); |
| 170 | + println!("Recall for k: {}: {:?}", k, recall); |
| 171 | +} |
| 172 | + |
| 173 | +#[tokio::main] |
| 174 | +async fn main() -> io::Result<()> { |
| 175 | + let base_folder = "examples/data/siftsmall"; |
| 176 | + let start_cpu = ProcessTime::now(); |
| 177 | + let start_wall = SystemTime::now(); |
| 178 | + |
| 179 | + let (base, query, ground_truth) = tokio::try_join!( |
| 180 | + SiftDataset::from_fvecs(format!("{}/siftsmall_base.fvecs", base_folder)), |
| 181 | + SiftDataset::from_fvecs(format!("{}/siftsmall_query.fvecs", base_folder)), |
| 182 | + SiftDataset::from_ivecs(format!("{}/siftsmall_groundtruth.ivecs", base_folder)) |
| 183 | + )?; |
| 184 | + |
| 185 | + println!("Base dataset: {} vectors of dimension {}", base.vectors.len(), base.dimension); |
| 186 | + println!("Query dataset: {} vectors of dimension {}", query.vectors.len(), query.dimension); |
| 187 | + println!("Ground truth: {} queries", ground_truth.len()); |
| 188 | + |
| 189 | + let index_name = "sift10k"; |
| 190 | + let index_view = IndexView::new(index_name).await.expect("Failed to create IndexView"); |
| 191 | + |
| 192 | + let path = Path::new(index_name); |
| 193 | + |
| 194 | + if !path.exists() { |
| 195 | + for (index, vector) in base.vectors.iter().enumerate() { |
| 196 | + index_view.insert(&VectorPoint::new(index as u32, vector.clone())).await.unwrap(); |
| 197 | + } |
| 198 | + } |
| 199 | + |
| 200 | + let index_cpu_time = start_cpu.elapsed(); |
| 201 | + let index_wall_time = start_wall.elapsed().unwrap(); |
| 202 | + println!("Indexing time: CPU {:?}, Wall {:?}", index_cpu_time, index_wall_time); |
| 203 | + |
| 204 | + for k in [1, 10, 100] { |
| 205 | + compute_recall_at_k(&index_view, &query, &ground_truth, k).await; |
| 206 | + } |
| 207 | + |
| 208 | + let cpu_time: Duration = start_cpu.elapsed(); |
| 209 | + let wall_time = start_wall.elapsed().unwrap(); |
| 210 | + |
| 211 | + println!("CPU time: {:?}", cpu_time); |
| 212 | + println!("Wall time: {:?}", wall_time); |
| 213 | + |
| 214 | + Ok(()) |
| 215 | +} |
0 commit comments