How to use multiple prefix tuned models with parallel inference.

[santhoshkolloju]

Hi Library error when I use muliptle prefix tuned models for parallel inference. What are the adpaters which supports parallel inference and how to know their configuration.

Thanks

[calpt]

Hey, parallel inference (as well as all other adapter composition blocks) are currently only supported with the "original" bottleneck adapter configuration (all adapters using AdapterConfig). Support for other methods such as prefix tuning may be added in the future.