File size: 12,538 Bytes
1269259 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 | // SPF Smart Gateway - BPE Tokenizer
// Copyright 2026 Joseph Stone - All Rights Reserved
//
// Byte-Pair Encoding tokenizer for SPF Transformer.
// Trains on SPF corpus (brain data, source code, rules).
// Pure Rust. No external tokenizer dependencies.
//
// Depends on: nothing (Layer 0)
use std::collections::HashMap;
// ============================================================================
// SPECIAL TOKENS
// ============================================================================
pub const PAD_TOKEN: &str = "[PAD]";
pub const BOS_TOKEN: &str = "[BOS]";
pub const EOS_TOKEN: &str = "[EOS]";
pub const UNK_TOKEN: &str = "[UNK]";
pub const TOOL_TOKEN: &str = "[TOOL]";
pub const GATE_TOKEN: &str = "[GATE]";
pub const USER_TOKEN: &str = "[USER]";
pub const SPF_TOKEN: &str = "[SPF]";
pub const ALLOWED_TOKEN: &str = "[ALLOWED]";
pub const BLOCKED_TOKEN: &str = "[BLOCKED]";
pub const PAD_ID: u32 = 0;
pub const BOS_ID: u32 = 1;
pub const EOS_ID: u32 = 2;
pub const UNK_ID: u32 = 3;
pub const TOOL_ID: u32 = 4;
pub const GATE_ID: u32 = 5;
pub const USER_ID: u32 = 6;
pub const SPF_ID: u32 = 7;
pub const ALLOWED_ID: u32 = 8;
pub const BLOCKED_ID: u32 = 9;
const NUM_SPECIAL: u32 = 10;
// ============================================================================
// BPE TOKENIZER
// ============================================================================
/// Byte-Pair Encoding tokenizer with SPF-specific special tokens
#[derive(Clone)]
pub struct Tokenizer {
/// Token string → ID
token_to_id: HashMap<String, u32>,
/// ID → token string
id_to_token: HashMap<u32, String>,
/// Merge rules: (pair) → merged token, ordered by priority
merges: Vec<(String, String)>,
/// Vocabulary size
pub vocab_size: u32,
}
impl Tokenizer {
/// Create a new tokenizer with only special tokens (untrained)
pub fn new() -> Self {
let mut tok = Self {
token_to_id: HashMap::new(),
id_to_token: HashMap::new(),
merges: Vec::new(),
vocab_size: NUM_SPECIAL,
};
// Register special tokens
let specials = [
PAD_TOKEN, BOS_TOKEN, EOS_TOKEN, UNK_TOKEN, TOOL_TOKEN,
GATE_TOKEN, USER_TOKEN, SPF_TOKEN, ALLOWED_TOKEN, BLOCKED_TOKEN,
];
for (i, &s) in specials.iter().enumerate() {
tok.token_to_id.insert(s.to_string(), i as u32);
tok.id_to_token.insert(i as u32, s.to_string());
}
tok
}
/// Load tokenizer vocabulary from JSON file.
/// Falls back to untrained tokenizer if file doesn't exist.
pub fn load(path: &str) -> Result<Self, String> {
let file_path = std::path::Path::new(path);
if !file_path.exists() {
// No trained tokenizer yet — use untrained (special tokens + byte fallback)
return Ok(Self::new());
}
let content = std::fs::read_to_string(file_path)
.map_err(|e| format!("Read error: {}", e))?;
Self::from_json(&content)
}
/// Save tokenizer vocabulary to JSON file.
pub fn save(&self, path: &str) -> Result<(), String> {
let json = self.to_json();
std::fs::write(path, json)
.map_err(|e| format!("Write error: {}", e))
}
/// Train BPE on a corpus of text. Learns `num_merges` merge rules.
/// `target_vocab` is the desired vocabulary size (including special tokens).
pub fn train(&mut self, corpus: &str, target_vocab: u32) {
let num_merges = target_vocab.saturating_sub(NUM_SPECIAL + 256); // 256 byte-level tokens
// Step 1: Initialize vocabulary with all byte values
for b in 0u8..=255 {
let token = format!("{}", b as char);
let id = NUM_SPECIAL + b as u32;
self.token_to_id.insert(token.clone(), id);
self.id_to_token.insert(id, token);
}
self.vocab_size = NUM_SPECIAL + 256;
// Step 2: Tokenize corpus into byte-level tokens
let mut words: Vec<Vec<String>> = corpus
.split_whitespace()
.map(|w| w.chars().map(|c| c.to_string()).collect())
.collect();
// Step 3: Iteratively merge most frequent pairs
for _ in 0..num_merges {
// Count all adjacent pairs
let mut pair_counts: HashMap<(String, String), u64> = HashMap::new();
for word in &words {
for pair in word.windows(2) {
*pair_counts
.entry((pair[0].clone(), pair[1].clone()))
.or_insert(0) += 1;
}
}
if pair_counts.is_empty() {
break;
}
// Find most frequent pair
let best_pair = pair_counts
.iter()
.max_by_key(|(_, &count)| count)
.map(|(pair, _)| pair.clone());
let (left, right) = match best_pair {
Some(p) => p,
None => break,
};
// Create merged token
let merged = format!("{}{}", left, right);
// Register in vocabulary
let id = self.vocab_size;
self.token_to_id.insert(merged.clone(), id);
self.id_to_token.insert(id, merged.clone());
self.merges.push((left.clone(), right.clone()));
self.vocab_size += 1;
// Apply merge to all words
for word in &mut words {
let mut i = 0;
while i + 1 < word.len() {
if word[i] == left && word[i + 1] == right {
word[i] = merged.clone();
word.remove(i + 1);
} else {
i += 1;
}
}
}
}
}
/// Encode text to token IDs
pub fn encode(&self, text: &str) -> Vec<u32> {
let mut tokens = Vec::new();
for word in text.split_whitespace() {
// Check if it's a special token
if let Some(&id) = self.token_to_id.get(word) {
tokens.push(id);
continue;
}
// Byte-level tokenization
let mut chars: Vec<String> = word.chars().map(|c| c.to_string()).collect();
// Apply merges in order
for (left, right) in &self.merges {
let merged = format!("{}{}", left, right);
let mut i = 0;
while i + 1 < chars.len() {
if chars[i] == *left && chars[i + 1] == *right {
chars[i] = merged.clone();
chars.remove(i + 1);
} else {
i += 1;
}
}
}
// Convert to IDs
for ch in &chars {
match self.token_to_id.get(ch) {
Some(&id) => tokens.push(id),
None => tokens.push(UNK_ID),
}
}
}
tokens
}
/// Decode token IDs back to text
pub fn decode(&self, ids: &[u32]) -> String {
let mut parts = Vec::new();
for &id in ids {
match self.id_to_token.get(&id) {
Some(token) => {
// Skip special tokens in output (except content ones)
match id {
PAD_ID | BOS_ID | EOS_ID => continue,
_ => parts.push(token.clone()),
}
}
None => parts.push(UNK_TOKEN.to_string()),
}
}
parts.join("")
}
/// Encode with BOS/EOS wrapping
pub fn encode_with_special(&self, text: &str) -> Vec<u32> {
let mut ids = vec![BOS_ID];
ids.extend(self.encode(text));
ids.push(EOS_ID);
ids
}
/// Get token string for an ID
pub fn get_token(&self, id: u32) -> Option<&str> {
self.id_to_token.get(&id).map(|s| s.as_str())
}
/// Get ID for a token string
pub fn get_id(&self, token: &str) -> Option<u32> {
self.token_to_id.get(token).copied()
}
/// Serialize tokenizer state to JSON
pub fn to_json(&self) -> String {
let merges_json: Vec<String> = self.merges
.iter()
.map(|(l, r)| format!("[\"{}\",\"{}\"]", l.replace('"', "\\\""), r.replace('"', "\\\"")))
.collect();
let vocab_json: Vec<String> = self.token_to_id
.iter()
.map(|(k, v)| format!("\"{}\":{}", k.replace('"', "\\\""), v))
.collect();
format!(
"{{\"vocab_size\":{},\"merges\":[{}],\"vocab\":{{{}}}}}",
self.vocab_size,
merges_json.join(","),
vocab_json.join(",")
)
}
/// Deserialize tokenizer from JSON string
pub fn from_json(json: &str) -> Result<Self, String> {
let parsed: serde_json::Value = serde_json::from_str(json)
.map_err(|e| format!("Tokenizer JSON parse error: {}", e))?;
let mut tok = Self::new();
// Load vocab
if let Some(vocab) = parsed.get("vocab").and_then(|v| v.as_object()) {
for (token, id) in vocab {
if let Some(id_num) = id.as_u64() {
let id32 = id_num as u32;
tok.token_to_id.insert(token.clone(), id32);
tok.id_to_token.insert(id32, token.clone());
}
}
}
// Load merges
if let Some(merges) = parsed.get("merges").and_then(|m| m.as_array()) {
for pair in merges {
if let Some(arr) = pair.as_array() {
if arr.len() == 2 {
if let (Some(l), Some(r)) = (arr[0].as_str(), arr[1].as_str()) {
tok.merges.push((l.to_string(), r.to_string()));
}
}
}
}
}
// Load vocab size
if let Some(vs) = parsed.get("vocab_size").and_then(|v| v.as_u64()) {
tok.vocab_size = vs as u32;
} else {
tok.vocab_size = tok.token_to_id.len() as u32;
}
Ok(tok)
}
}
// ============================================================================
// TESTS
// ============================================================================
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_special_tokens() {
let tok = Tokenizer::new();
assert_eq!(tok.get_id(PAD_TOKEN), Some(PAD_ID));
assert_eq!(tok.get_id(BOS_TOKEN), Some(BOS_ID));
assert_eq!(tok.get_id(TOOL_TOKEN), Some(TOOL_ID));
assert_eq!(tok.get_id(GATE_TOKEN), Some(GATE_ID));
assert_eq!(tok.get_token(SPF_ID), Some(SPF_TOKEN));
}
#[test]
fn test_train_and_encode() {
let mut tok = Tokenizer::new();
let corpus = "the cat sat on the mat the cat the mat";
tok.train(corpus, 300);
let ids = tok.encode("the cat");
assert!(!ids.is_empty());
assert!(ids.iter().all(|&id| id != UNK_ID));
}
#[test]
fn test_roundtrip() {
let mut tok = Tokenizer::new();
tok.train("hello world hello world", 300);
let text = "hello";
let ids = tok.encode(text);
let decoded = tok.decode(&ids);
assert_eq!(decoded, text);
}
#[test]
fn test_encode_with_special() {
let mut tok = Tokenizer::new();
tok.train("test data", 300);
let ids = tok.encode_with_special("test");
assert_eq!(ids[0], BOS_ID);
assert_eq!(*ids.last().unwrap(), EOS_ID);
}
#[test]
fn test_unknown_token() {
let tok = Tokenizer::new(); // untrained — only specials + no bytes
// With an untrained tokenizer, byte-level chars might not exist
// This tests the UNK fallback path
assert_eq!(tok.vocab_size, NUM_SPECIAL);
}
#[test]
fn test_json_roundtrip() {
let mut tok = Tokenizer::new();
tok.train("the cat sat on the mat", 300);
let json = tok.to_json();
let tok2 = Tokenizer::from_json(&json).unwrap();
assert_eq!(tok.vocab_size, tok2.vocab_size);
assert_eq!(tok.merges.len(), tok2.merges.len());
let ids1 = tok.encode("the cat");
let ids2 = tok2.encode("the cat");
assert_eq!(ids1, ids2);
}
}
|