Adding Crusoe examples to 3p_integrations

[ethxnp]

🚀 The feature, motivation and pitch

Prospective Crusoe customers would like to understand how to perform common Llama workflows (inference, fine-tuning, training) on our cloud. We'd like to contribute solutions centered around Llama to the 3p_integrations repo, beginning with a tutorial on benchmarking FP8 quants served via vLLM. The tutorial covers how to deploy resources on Crusoe, start a vLLM server, run and interpret benchmarks, and finally how to create fp8 quants of existing Llama3 finetunes.

We hope for this to be the first of a series of solutions for common Llama workflows!

Alternatives

No response

Additional context

No response

[init27]

Hi @ethxnp thanks for starting the chat!

Yes-this would be really cool to have-please feel free to start a PR and tag me if you have any Qs!

[ethxnp]

@init27 thank you! Just submitted a PR

[HamidShojanazeri]

Thansk @ethxnp for the PR