[router] Add Configurable L0 and L1 Tokenizer Caching (#11688)

2025-10-18 18:33:53 -07:00
parent fda0cb2a30
commit a7ae61ed77
22 changed files with 2385 additions and 24 deletions
--- a/sgl-router/src/tokenizer/tests.rs
+++ b/sgl-router/src/tokenizer/tests.rs
@@ -43,7 +43,7 @@ fn test_tokenizer_wrapper() {
    let text = tokenizer.decode(&[1, 2], false).unwrap();
    assert_eq!(text, "Hello world");

-    assert_eq!(tokenizer.vocab_size(), 8);
+    assert_eq!(tokenizer.vocab_size(), 14);

    assert_eq!(tokenizer.token_to_id("Hello"), Some(1));
    assert_eq!(tokenizer.token_to_id("unknown"), None);