pub struct UnknownPretokenizer {
pub pre: String,
pub spm_style: bool,
}Expand description
A model declared a pre-tokenizer memra has no exact split for.
Before 2026-08-19 this was one eprintln! per process followed by a silent fall-through to
the qwen35 split: the model loaded, generated fluent text, and every token id was wrong —
the same fluent-and-invisible class as the GGUF chat-template mint trap. Wrong ids poison
goldens, parity fixtures, acceptance counts and every quality number downstream, and nothing
in the stack can detect it after the fact. So it is a hard load error now.
Fields§
§pre: StringThe value that was rejected (tokenizer.ggml.pre, or default when an HF checkpoint’s
pre-tokenizer regexes matched no known family).
spm_style: boolTrue when the vocab model is SPM-style (tokenizer.ggml.model == "gemma4"); a pre/model
disagreement is itself the fault, so it is worth naming.
Trait Implementations§
Source§impl Clone for UnknownPretokenizer
impl Clone for UnknownPretokenizer
Source§fn clone(&self) -> UnknownPretokenizer
fn clone(&self) -> UnknownPretokenizer
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
source. Read moreSource§impl Debug for UnknownPretokenizer
impl Debug for UnknownPretokenizer
Source§impl Display for UnknownPretokenizer
impl Display for UnknownPretokenizer
impl Eq for UnknownPretokenizer
Source§impl Error for UnknownPretokenizer
impl Error for UnknownPretokenizer
1.30.0 · Source§fn source(&self) -> Option<&(dyn Error + 'static)>
fn source(&self) -> Option<&(dyn Error + 'static)>
1.0.0 · Source§fn description(&self) -> &str
fn description(&self) -> &str
use the Display impl or to_string()