[Feature]: Support extendable configuration files #24096
Description
🚀 The feature, motivation and pitch
It's common to have different configuration files that essentially serve the same system with minor changes (i.e. using a different model or just changing the topology).
While it is possible to provide these changes in the command line arguments, it would be nice to have the support for extending configuration files, so that you could have:
"serve_azure.yml" - base config
"serve_azure_qwen32.yml" - extends base with qwen32 model
"serve_azure_Llama-3.2.yml" - extends base with Llama-3.2 model
And instead of having to copy the same configurations both could have some key like "extends" to reference the base config and only change what is required.
Alternatives
The alternative solution could be the ability to specify multiple configuration files in serving args and merge their dictionaries, but then it might be a bit confusing when looking at the configuration files alone, or given the fact that the order for providing them will change the behavior ...
Additional context
No response
Before submitting a new issue...
- Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.