Expand description
Explicit developer/lab admission of one Qwen3.5 linear layer. This is not whole-model decode admission. Packing and ordinary norm storage are caller-selected source contracts; no exporter-version heuristic is used.
Modules§
- ftw
- Architecture-owned, explicitly admitted FreeToken Qwen3.5 text execution. The generic GGUF/Llama admission remains unchanged.
Structs§
- Prepared
Linear Layer - Cold owned execution buffers plus retained exact source owners. Dropping or replacing the engine cannot redirect these descriptors to another model. Only validated native prepared owners or explicit immutable leases enter.
- Profile