Best way to run (or otherwise just run) an LLM inside colab instance and using it through litelmm?

[cheery]

Hi,

I'm particularly interested about using semlib, but semlib uses litellm under. I'd need structured outputs.

What would be the most convenient way to use an LLM through litellm inside the colab instance? I want to utilize colab's GPU to run the LLM.

I found instructions to run ollama inside colab, but this seems very slow to setup. Either way. The most convenient way to get this semlib -library running would be appreciated.