Ability to choose what model to use for data synthesizer

[INF800]

Instead of simply allowing data_synthesis=True while finetuning, we can do something like:

data_synthesis = True      // uses gpt-4 model
data_synthesis = 'gpt-4'      // uses gpt-4 model
data_synthesis = 'gpt-3.5-turbo;      // uses chatgpt model
data_synthesis = 'random'      // uses models in random

[INF800]

Needed this because my API key did not have access to gpt4 and there wasn't any direct way for changing data synthesis model

[mmirman]

Before we can fix this we need to:

1. make sure the model can do data synthesis, and set it up
2. document which models are good at what with respect to data synthesis.
3. Figure out how to measure these things!

[lucylililiwang]

Hi, can I please work on this issue? Thank you!

[lucylililiwang]

https://github.com/lucylililiwang/LLM-VM/blob/lucylililiwang-choose_model/choose_model.py