Add another SUT or two

[wpietri]

Yifan's suggestions:

I'd suggest openai/gpt-3.5-turbo-0613 next (which is relatively inexpensive)
Oh, also EleutherAI/pythia-70m is also small enough to work on CPU
but it is not a built in model in HELM yet, so that forces us to think about how the SUT plugins will work, which is maybe good?
on the HELM CLI, you would run pythia-70m [like so](https://crfm-helm.readthedocs.io/en/latest/huggingface_models/)
GPU models: pythia, mpt, llama 2
llama 2 and the openai models also need credentials, so that also forces us to think about how credentials should be managed

[yifanmai]

Open source Hugging Face SUTs:

• gpt2
• EleutherAI/pythia-70m
• facebook/opt-125m

Closed source:

• gpt-3.5-turbo-1106

[yifanmai]

Open source remote inference projects that use the OpenAI API format:

• vLLM
• OpenLLM