pdserving multi-model deployment

[vinnyt123]

I'm wondering if it is possible to deploy multiple language models at once with pdserving?

Ideally I would like to have multiple recognition inference models available so I can conditionally do recognition based on a language parameter.

I see the inference model location is defined in in the config.yml. Is it possible to change this location at runtime? I am using the python mode for deployment.

[tink2123]

There is no direct parameter to support the above behavior, but you can try to implement it like this:

1. start multiple server services
2. judge the distribution by customizing the language parameters on the client side.