pub fn run_decode_qwen4exp(
package: &Path,
prompt: Option<&str>,
token_ids: &[u32],
max_tokens: usize,
context: usize,
chat: bool,
trunk_mmap: bool,
) -> Result<(), String>Expand description
Real multi-token Qwen4Exp generation: prompt → token embedding → PLE →
all 48 trained layers (HC → GDN|QSA → HC → MoE) → final HC mixer →
streamed output.weight argmax → native tokenizer decode. Every step
streams rows from the C: PLE payload and E: trunk; no model residency.