Serve pretrained model as API

[masus04]

I am trying to start a server using a pretrained model. Am I correct in assuming the server.py is used to start an inference server?

As far as I understand the available models are configured in the /available_models/conf.json file. Is there any documentation on the options available for this config file?

[goncalomcorreia]

Hi, this is an old fork of OpenNMT-py. This seems like an issue specific to that repository. I never used the server.py script so I don't know if the changes in this fork have an effect on it in any way

[masus04]

Ok, thank you.
Is there any other existing method to serve a pretrained model?

[goncalomcorreia]

Not that I know of.

[goncalomcorreia]

You can try to see the documentation for server.py in the original OpenNMT-py repository. If you find the issue, I'll gladly accept a PR :)

[masus04]

Alright, will check, thank you 👍