The current vLLM version supported by kvcached lacks the right way of loading gemma4 models, e.g., gemma-4-26b-a4b. Maybe this will be fixed after new versions of engines are supported. Related to: https://github.com/ovg-project/kvcached/issues/358
The current vLLM version supported by kvcached lacks the right way of loading gemma4 models, e.g., gemma-4-26b-a4b.
Maybe this will be fixed after new versions of engines are supported. Related to: #358