Module common 

Common executor utilities — extracted from duplicated code across Qwen3, Qwen2, and Llama executors.

Structs

GenericKvCacheHandle

Generic KV cache handle usable by any model architecture.

Functions

default_executor_status

Default executor status (all executors return the same thing).

tensor_to_tokens

Extract token IDs from a TensorRef.

tokens_to_tensor

Convert token IDs to a candle Tensor on the target device.

wrap_tensor

Wrap a candle Tensor as a TensorRef.